# Glowstone guides > Step-by-step guides to every Glowstone feature, written for people and for AI agents. > Glowstone is a free macOS workspace that gives the user's own AI agent (Claude Code, Codex, Cursor, OpenCode, Kilo Code or any CLI agent) a memory: meetings, notes and decisions kept as plain Markdown files on the user's Mac. > Button names below are the app's own. When helping a user, quote them exactly. ## Meetings: A meeting assistant that knows your business. https://glowstone.ai/guides/meetings Record from the tab at the top of your screen, ask while the call is still going, and turn the transcript into notes with one click. No bot joins the call, and the audio never leaves your Mac. ### Why it matters - **It answers with more than the meeting.** A meeting recorder only knows what was said in the room. Your agent also reads your notes, your past meetings and the tools you connected (calendar, Stripe, Linear, your CRM). So “did we promise this customer anything?” gets a real answer, mid-call, not a summary afterwards. - **You stop taking notes.** Listen and talk. Everything is written down for you, split between what you said and what the other side said, so you can be in the conversation instead of typing about it. - **It files itself.** One click on Process and your agent goes through the transcript: decisions saved, people and companies updated, and what’s still pending written down, with who it’s pending on. - **You can ask it weeks later.** The transcript stays in your workspace as a note. “What exactly did Maya say about the redesign date?” comes back with the answer and the minute it was said. ### How it works 1. **Start recording from the tab.** Bring the pointer to the top edge of the screen and the tab opens. The small round button next to the agent’s name starts recording: the name turns into a running clock, the button into Stop, and Transcript and Chat appear. Move the pointer away and the tab tucks back into the edge while it keeps recording. Reach for it again and the transcript is there. Your side comes from the microphone and theirs from what your Mac is playing, so it knows who said what without anyone joining the call. Hearing the other side needs macOS 14.2 or newer. (Shown: Pointer to the top edge, the tab opens, one click on the round button; it tucks away when you leave and keeps recording.) 2. **Ask while the call is still going.** On the Transcript side the lines arrive a few seconds after they’re said. Type your question in the box at the bottom; when you send it, the tab switches to Chat. The answer here can’t come from the meeting alone. What you promised Northwind lives in a note from September, and your agent reads it while the call goes on. It doesn’t write anything into your workspace unless you ask it to. (Shown: Transcript first, then a question the meeting alone can’t answer.) 3. **Another call: your tools join in.** Same tab, a sales call this time. The question needs your price list and the customer’s account, so the agent reads Pricing.md and checks Stripe through the connector you set up once. This is the difference with a recorder: the context isn’t the meeting, it’s your business. (Shown: A connector (Stripe) and a note (Pricing.md), answered before they finish the sentence.) 4. **Stop, then Process.** Click Stop in the tab. Glowstone transcribes what’s left, saves the note in meetings/ and comes to the front on Meetings, with the call at the top of the list. Process, on the meeting’s row (or Ask the agent, inside the transcript), opens a new chat with the request already written: digest this meeting into the vault, save what was decided, update the people, companies and deals that came up, and note what’s pending and on whom. It doesn’t summarise it back at you: you were there. The request also warns the agent that an automatic transcript can get names and figures wrong, so before writing down an amount, a date or a commitment it isn’t sure of, it asks you. (Shown: Stop in the tab, the meeting lands in Meetings, Process sends it to your agent.) 5. **Ask about it any time later.** The transcript is a normal note, and your agent finds it on its own. Ask what someone said, what you promised, or have it draft the follow-up email in your words. (Shown: Two weeks later, from any chat.) ### Things to ask - “What did we just agree for this sprint?” - “Did we promise Northwind anything before their renewal?” - “Is this customer behind on any invoice?” - “What did I commit to in this call?” - “Draft the follow-up for the team from this meeting.” - “What changed since last sprint planning?” - “Who owns each action item, and by when?” ### Facts - `~10 s`: From someone speaking to the line on screen - `99`: Languages, detected from the audio - `0`: Bots in the call. Audio is never saved, only the text - `meetings/`: Where each transcript lands, as a Markdown note - `Free`: Transcription runs on your Mac with the included Whisper models ### Questions - **Does a bot join the call?** No. Glowstone listens on your Mac: your microphone for you, the system audio for everyone else. It works with any call app, because it never joins the call. - **Is the audio stored or uploaded?** No. Speech is transcribed by a model running on your Mac and the audio is discarded. What stays is the text, in a note you can read, edit or delete. Your questions go to the agent you chose, like any other chat. - **Does transcription cost anything?** No. Glowstone includes the transcription models and runs them on your Mac, so there’s no per-minute fee or meeting-recorder subscription. Questions you ask the agent use your own agent’s plan, as any chat does. - **Does it work for in-person meetings?** Yes. In a room everyone reaches the microphone, so the transcript can’t split voices into You and Them; it’s one track. Everything else works the same. - **Do I need to tell people?** Yes. Recording other people is regulated in many places. Tell them first. ## The tab: Your agent, at the top of your screen. https://glowstone.ai/guides/the-tab A small black tab hangs from the top edge of your screen in every app. It opens when you reach for it, and it has the same agent and the same memory as the main window. ### Why it matters - **No switching windows.** The quick question you would have forgotten by the time you opened another app gets asked right there, over whatever you’re working on. - **Same memory everywhere.** It isn’t a lighter assistant. The tab talks to your agent with your whole workspace behind it, and what you jot down there lands in the same notes. - **It can see what you see.** One button attaches what’s on screen. “What does this error mean?” or “add these numbers to the Northwind note” stop needing an explanation of what “this” is. ### How it works 1. **Reach for it.** At rest it’s a thin black notch at the top of the screen. Bring the pointer to it and it opens over whatever you’re doing; move away and it tucks back in. From the keyboard, ⌘⇧Space or ← and → together take you to the main Glowstone window instead. (Shown: The pointer reaches the top edge, the tab opens, and tucks back in when it leaves.) 2. **Jot it down before it’s gone.** Type the thought the way it came to you. Your agent decides where it belongs, links it to the right person or project, and you go back to what you were doing. (Shown: Up to the edge, a line, Enter. It’s filed.) 3. **Show it the screen.** The camera button attaches a screenshot of your screen to the message. Pressing ↓ and → together does it in one gesture: a new chat with the screenshot already attached. (Shown: The camera attaches what’s on screen, and the question can just say “this”.) ### Things to ask - “Note: Leo wants the Beacon test moved to Thursday.” - “What’s on my plate before the 11:00?” - “What does this error mean? (with the screen attached)” - “Add these numbers to the Northwind note.” - “Remind me what we decided about SSO.” - “Start recording: sprint review.” ### Shortcuts, set up your way The tab is the quick lane: move the pointer up and your agent is there, over whatever you’re doing. For the rest there are gestures, two keys pressed together from any app, and each one is yours to change. Pick its keys or switch it off in Settings → Gestures. - `← + →`: Bring Glowstone up. From any app, to the main window. - `↓ + →`: New chat with what’s on screen. A screenshot, already attached. - `↑ + ↓`: Start or stop dictating. Or hold 🌐 fn, or the right ⌘: you choose the key. - `⌘ + ⇧ + Space`: Open the Glowstone window. A classic shortcut, always on. Gestures ship switched off and ask for Accessibility the first time, because reading two keys pressed together in another app is what that macOS permission is for. In Settings → Gestures each row lights up when you do the gesture, so you know it works. ### Questions - **Can I record a meeting from the tab?** Yes, that’s where it starts. See the Meetings guide. ## Dictation: Dictation that’s free, and never leaves your Mac. https://glowstone.ai/guides/dictation Hold a key, speak, let go. The text is typed wherever your cursor is, usually in about a second, in any app. The speech model is Whisper, included with Glowstone and running on your Mac: no dictation subscription, and your voice never goes to anyone’s server. ### Why it matters - **No subscription for your voice.** Most dictation apps charge every month and send what you say to their servers to turn it into text. Glowstone includes the transcription models: you download one once and it’s yours. Dictating and transcribing meetings cost nothing, however much you talk. - **You say more than you type.** Speaking is about three times faster than typing, and people say more when they talk: the context, the why, the detail that makes an agent’s answer good. - **Your Mac does the listening.** The transcription happens on your own machine. It works with no connection, and what you say isn’t uploaded, stored or used by anyone to train anything. The audio is discarded once it’s text. - **It knows the names of your work.** A generic transcriber hears “Vercel” and writes “Versel”. Glowstone reads the names in your workspace (people, companies, projects) and learns the ones you correct, without ever swapping an ordinary word for one of them. ### How it works 1. **Turn it on once.** Settings → Voice. It tells you the price before charging it: the size of the model it will download, the microphone permission, and Accessibility, which is what lets it type into other apps. You choose the key during onboarding: 🌐 fn, the right ⌘, or two arrow keys pressed together. You can change it any time in Settings → Gestures, along with the rest of the shortcuts. 2. **Hold, talk, let go.** Hold the key in Mail, Slack, a document or a chat with your agent. A small pill shows it’s listening; when you let go, the text appears where the cursor was. Hold it and say nothing, and nothing is typed. For long stretches, hold fn and tap space to leave it running, then space again to send. With the right ⌘ it’s one press to start and one to stop; with two arrows, the same chord starts and stops it. (Shown: In any app: hold, talk, let go.) 3. **Correct a name once.** Names that already have a note go in on their own. For the rest (tools, products, a surname that’s written nowhere) tell your agent “it writes Versel and it’s Vercel” and it adds the word to your list in .glowstone/vocabulario.md. (Shown: Corrected once, spelled right from then on.) ### Things to ask - “Dictate the weekly update for the Atlas team.” - “Reply to Sophie with the export date.” - “Brain-dump the Q4 launch risks into a note.” - “Talk a spec through out loud, then ask the agent to tidy it.” - “Answer a long Slack thread without typing it.” ### Facts - `Free`: Whisper models included. No subscription, no word limit - `~1 s`: From letting go to text, on a MacBook Air - `99`: Languages, detected on their own - `Offline`: Runs on your Mac, no connection needed ### Questions - **Is it really free?** Yes. Glowstone is free, and dictation doesn’t add a subscription: the speech models are open (Whisper, through whisper.cpp) and run on your Mac, so there is nothing to meter. The only thing you might pay for is the AI agent you chat with, on your own plan, and dictation doesn’t use it. - **Does my voice go anywhere?** No. It’s transcribed on your Mac and the audio is never saved; only the text is typed where your cursor was. Glowstone keeps your last 50 dictations as text, on your Mac, so you can copy one again, and you can clear them in Settings → Voice. If you then send that text to your agent, it goes wherever your agent goes, like anything you type. - **Which model should I use?** The default (base) is small and fast. If you work with strong accents or a less common language, Settings → Voice offers larger Whisper models (large-v3-turbo and large-v3) that are slower and more precise, with their size shown before you download. ## Memory: Notes you don’t have to write. https://glowstone.ai/guides/memory Your workspace is a folder of plain Markdown files on your Mac: one note per person, company, project, meeting and decision. Your agent writes and links them while you work, and reads them before it answers. ### Why it matters - **It fills itself.** Second brains fail because filing is a second job. Here the meetings, the dictation and the agent fill it, so it stays current without discipline. - **Every answer starts from what you know.** The agent reads the relevant notes before answering, so “did we promise SSO?” is answered from the decision you took, not guessed. - **It’s just files.** Open them in Obsidian, a text editor or Finder. Back them up by putting the folder in iCloud Drive, Google Drive or Dropbox. Change agent tomorrow and the memory comes with you. ### How it works 1. **Your notes, inside Glowstone.** Switch the right column from Memory to Documents and the whole vault is there: one folder per kind of thing, the notes your agent wrote, and anything you added yourself. Open a note to read or edit it in place, with its links to everything else. A folder appears when its first note arrives, never before. Every note has a short header (type, status, who’s involved), which is what lets the agent answer questions instead of only searching text. (Shown: Documents, a folder, a note: the vault from inside Glowstone.) 2. **And it’s a folder on your Mac.** What Glowstone shows you is a normal folder on your disk: people/, companies/, projects/, meetings/, decisions/, daily/. Open it in Finder, Obsidian or any text editor and you find the same files with the same names. Nothing lives in a database you can’t open. (Shown: The same folder in Finder: what you see is what the agent reads.) 3. **Your agent writes as you talk.** Mention that Maya’s export passed QA and the agent updates Atlas.md, Maya’s note and today’s. Every change is shown under the answer, and you can review it line by line and undo it. (Shown: A sentence in the chat, three notes updated, every change reviewable.) 4. **Links make the graph.** Notes mention each other with [[links]]. That’s what lets you ask “what do I know about Marta?” or “what does the data team have open?”, and what the graph on the right draws. It lights up as the agent reads. (Shown: Notes arrive, links connect them, and the ones the agent reads light up.) ### Things to ask - “What do I know about Northwind?” - “What’s open on Atlas, and who is it waiting on?” - “When did we decide export goes before the redesign, and why?” - “What has Maya committed to this month?” - “Summarise Beacon’s user tests so far.” - “Make a note for every person in today’s meetings.” ### Your memory is yours. Not ours. Glowstone is an app on your Mac, not a cloud service. Your notes are a folder on your disk and your chats are stored on your Mac. There’s no Glowstone account, and no Glowstone server that receives any of it. - **We can’t read it.** Not your notes, not your chats, not your meetings. Nothing is uploaded to us, so there is nothing for us to read, lose or hand over. On its own the app only checks for updates and downloads the speech model once. - **Only the agent you choose sees it.** When you ask something, your agent reads the notes it needs to answer: Claude Code sends that to Anthropic, Codex to OpenAI, under your own account and their terms. Choosing the agent is choosing who sees your work, and that’s your call, not ours. - **Take it anywhere, any time.** It’s plain Markdown in a folder. Open it in Obsidian, put it in iCloud or Drive, copy it to another Mac. No export step, no lock-in, nothing to cancel: stop using Glowstone tomorrow and every note is still yours, where it was. Two things only leave your Mac if you set them up: connectors (they belong to your agent) and Telegram (messages go through Telegram). And if you ever send us a problem report, you see everything it contains before it’s sent. ### Questions - **Can I use my Obsidian vault?** Yes. Glowstone uses plain Markdown and wiki-style links, so you can point it at your vault or have it copied. See the guide on bringing your memory. - **Is there sync between Macs?** No. The workspace is a folder, so put it inside iCloud Drive, Google Drive or Dropbox if you want it everywhere, and point Glowstone at it in Settings → Change folder. ## Routines: Some conversations should start themselves. https://glowstone.ai/guides/routines Ask your agent, in words, to check in every morning, prepare you before each meeting or review the week on Fridays. When the time comes, it starts the conversation. When there’s nothing to say, it stays quiet. ### Why it matters - **The follow-up you would have skipped.** Reviewing what’s open is the habit everyone means to have. A routine turns it into something that arrives, already done, at the time you chose. - **Silent when there’s nothing.** A routine that checks your calendar every 15 minutes is only bearable because it speaks only when something is about to happen. - **No form to fill.** You say “every weekday at nine”; the agent writes it in .glowstone/rutinas.md, a normal file you can read and edit, and repeats back what it understood. ### How it works 1. **Say what and when.** In any chat. The agent turns “every morning at nine” into the schedule, writes the routine and tells you, in your words, what it understood. If what you ask doesn’t fit the schedules it supports, it says so instead of approximating. (Shown: A sentence becomes a routine in a file you can open.) 2. **It opens the conversation.** At the time, the agent runs against your workspace and its answer arrives as a new chat in your list, with a notification. You can reply to it like any other. Routines run while Glowstone is open. They don’t wake a sleeping Mac. (Shown: 09:00: the conversation starts without you.) ### Things to ask - “Every weekday at nine, tell me what’s still open.” - “Before every meeting, tell me who’s coming and what we left open.” - “Every Friday at five, review my week.” - “On the 1st of each month, list invoices that are still unpaid.” ### Facts - `daily 09:00`: Every day - `weekdays 09:00`: Monday to Friday - `weekly fri 17:00`: One day of the week - `monthly 1 09:00`: One day of the month - `once 2026-10-20 16:00`: A single time - `every 15m`: Every so often (5 minutes minimum) ## Connectors: Your agent, plugged into your tools. https://glowstone.ai/guides/connectors Calendar, Gmail, Linear, Notion, Stripe, Sentry, GitHub. Tell your agent what you use and it connects them, right in the conversation. Then every chat, meeting and routine can use them. ### Why it matters - **From what you tell it to what’s true.** Without connectors your agent knows what’s in your notes. With your calendar connected it can prepare a meeting before you walk into it; with Stripe, tell you a customer is two invoices behind while you’re on the phone. - **Yours, not ours.** Connectors are your own agent’s MCP servers. What you connect in Glowstone also works in your Terminal, and the other way round. There’s no Glowstone account in the middle. ### How it works 1. **Say what you use.** Start from your day: where your calendar, email, tickets and numbers live. For each tool the agent looks for the official connector, adds it with your agent’s own command, and keeps a table in My tools.md: connected, not connected, or no connector available, with the date it checked. If you say no to one, it writes that down too, and doesn’t ask again. (Shown: Two tools named, two connectors added, one table that remembers.) 2. **Sign in once.** Most connectors open the browser for you to authorise them. The chat header tells you when one is waiting to be signed in. They load when a conversation starts, so use the new one from a new chat. A few, like GitHub and Slack, use a token instead of a browser sign-in. Your agent walks you through it, and the token never goes into a note. 3. **Then it’s everywhere.** In the tab before a call, inside a live meeting, in a routine at 9:00. The same connector serves all of them. (Shown: Calendar and notes together, ten minutes before the call.) ### Things to ask - “What do I need for my 11:00?” - “Which Linear issues block the Atlas export?” - “Is Northwind up to date on payments?” - “What broke in Sentry since yesterday’s Beacon release?” - “Create a Linear ticket from this meeting’s action items.” ### Facts - `Linear`: What you’re working on, and with whom - `Notion`: The notes you already had elsewhere - `Stripe`: Customers, invoices, what people pay - `Sentry`: What’s breaking before someone says so ## Bring your memory: Don’t start from a blank page. https://glowstone.ai/guides/bring-your-memory Bring what ChatGPT or Claude already know about you, your Claude Code and Codex conversations, or an Obsidian vault. Your agent turns it into notes on day one. ### Why it matters - **Useful from the first question.** A memory is only worth something once it has something in it. Months of context you already gave another assistant shouldn’t stay locked there. - **Copied, never moved.** Imports read and copy. Your Obsidian vault, your chats and your ChatGPT account stay exactly as they were. ### How it works 1. **From ChatGPT or Claude.** Three steps: copy a request Glowstone gives you, ask it in ChatGPT or Claude, paste the answer back. Your agent turns it into notes: who you are, your work, the people and projects that came up. It’s a step in onboarding, and it’s always under the archive icon. (Shown: Copy, ask there, paste here.) 2. **Your Claude Code and Codex chats.** Pick which conversations to bring, grouped by project. This week’s come checked. They arrive with their dates, their project as a tag, and archived if they were archived. (Shown: Grouped by project, so dropping a whole one is one click.) 3. **An Obsidian vault or a Markdown folder.** Point Glowstone at it, or ask your agent to copy it into your workspace, keeping what it already had. If it’s an export from somewhere else, the agent reads it and creates the notes and people that come out of it. ### Things to ask - “What did ChatGPT know about how I like to work?” - “Find the Claude Code chat about the usage query.” - “Which of my old chats mention Northwind?” - “Turn my old Obsidian notes on Beacon into project notes.” ## Search & spaces: Find it by what was said. https://glowstone.ai/guides/search Search your notes and your conversations by what’s inside them, tag the threads you keep coming back to, and keep work and home in separate spaces that never mix. ### Why it matters - **You remember a phrase, not a title.** Search looks inside notes and inside conversations, ignoring accents and case, so the half-sentence you remember is enough. - **When search fails, ask.** The last row of every search is “ask the agent” with what you typed. The memory and what reads it are the same place. ### How it works 1. **⌘F, anywhere in the app.** It searches titles and the inside of notes. If what you’re after isn’t there, the last row sends what you typed to your agent, which can read between the lines search can’t. (Shown: Results as you type, and the agent as the last row.) 2. **Conversations: search and tags.** The search above the chat list looks inside what was said and shows the matching bit. Tags come in colour from the start, a thread can carry several, and filters add up. (Shown: A phrase finds the thread; a tag narrows the list.) 3. **Spaces for different parts of your life.** Each space is its own folder with its own conversations. Adding something to the shopping list at home doesn’t touch work, and the agent in one never sees the other. (Shown: Two folders, two memories, one click between them.) ### Things to ask - “Where did we write down the export promise?” - “Show me every client thread from this month.” - “What did I ask about SSO last week?” - “Which meeting was the redesign date set in?” ## Telegram: Your workspace, from your phone. https://glowstone.ai/guides/telegram Write or send a voice note to your own Telegram bot to jot something down or ask about your notes. Your agent answers from your Mac, with the same memory. ### Why it matters - **The thought you have away from the desk.** Leaving a viewing, after a call in the car, in a queue. Say it into Telegram and it’s in the right note by the time you’re back. - **Private by construction.** It’s your bot, paired to your account. The token lives in the macOS Keychain and never passes through a model, and voice notes are transcribed on your Mac. ### How it works 1. **Create your bot and pair it.** Settings → Telegram walks you through it: create a bot, paste its token, then pair by scanning a QR code or typing the number the bot gives you. Pairing is what stops anyone who finds the bot from talking to your workspace. 2. **Write or talk.** Text and voice notes up to 10 minutes. Each Telegram chat keeps one thread going; /new starts another and /threads lists the recent ones. After three hours of silence the next message starts fresh. What mattered is already in your notes. Every thread shows up in the sidebar on your Mac, so you can carry on there. It answers while Glowstone is open on your Mac. (Shown: A voice note on the way out, filed before you’re back.) ### Things to ask - “Note: Beacon onboarding step 3 confuses users.” - “What’s left for Friday?” - “Who’s in my 3 pm?” - “Add to Atlas: Maya is out on Monday.” # Glowstone for your work > How Glowstone helps specific professions. Same features, applied to each job. ## The AI second brain for consultants https://glowstone.ai/for/consultants Every client, remembered. Glowstone is like an assistant who has read everything about your practice: every client, every meeting, every decision and every promise. It thinks with the judgment of your sharpest colleague and answers from your own records, so you walk into every call already briefed. ### The problem - **The context lives in your head.** Who pushed back on what, which version they saw, what you promised on the last call. Switch clients five times a day and some of it falls out. - **Notes you never finish.** You’re either listening or typing. Write-ups pile up for Friday, and by Friday half the nuance is gone. - **Follow-ups that slip.** “I’ll send that over by Thursday” is said in passing and lives nowhere. The client remembers. You find out when they ask. ### Keep every client on record and ask your AI anything about them Each client gets its own notes: the engagement, the people and what they care about, every meeting, every decision and what you’ve promised. Your agent writes and links them as you work, so they stay current without a filing session. Then you ask it anything, in plain words, the way you’d ask a colleague who never forgets: what the CFO objected to, when a scope change was agreed, what’s still open before the next steering committee. - One note per client, person, meeting and deliverable - Decisions saved with the why and the date - Answers that cite the note they came from (Shown: A client’s notes inside Glowstone, then a question answered from them.) Example. It’s Thursday afternoon. Acme’s CFO has a steering committee tomorrow and you haven’t touched that engagement in nine days. You ask: “Brief me on Acme: where we are, what Hannah objected to, and what I owe her.” Glowstone: Reads the engagement note, the last two meetings and Hannah’s note; Pulls out the open objection and the promise you made; Checks deliverables/ for the draft you started. Result: A five-line brief: status, Hannah’s churn objection, the revised model due Friday and where the draft is. Two minutes instead of thirty. ### Record client meetings and ask questions live, with the meeting and the client’s history as context Start recording from the tab at the top of your screen. No bot joins the call, so nothing changes for the client: Glowstone hears your microphone and what your Mac plays, and transcribes it on your Mac. When a question comes up you can’t answer from memory, ask the tab mid-call. The answer comes from the conversation so far and from everything you already know about that client. Afterwards, Process files the decisions, the people and the follow-ups into the right notes. - No bot in the call, nothing for the client to accept - Ask mid-call, with the client’s history behind the answer - Process turns the transcript into decisions and follow-ups (Shown: A steering committee: the client asks, you ask the tab, the answer comes from the September workshop.) Example. Mid steering committee, the CFO asks a number from a workshop three weeks ago. You ask: “What did the September workshop say about churn by region?” Glowstone: Reads the transcript of the call so far; Opens the Sep 18 workshop notes; Answers with the figure and where it came from. Result: You answer on the spot with 6.1% vs 3.4%, instead of “I’ll get back to you”. ### Use it all day: capture what’s on screen, jot notes and ask from any app Most of the value comes from small moments, dozens of times a day. Glowstone hangs from the top of your screen in every app: bring the pointer up, ask, jot something down or attach what’s on screen, and get back to work. Away from the desk, a voice note to your own Glowstone bot on Telegram does the same. Everything lands in the same memory. - The tab, over any app - Screenshots it understands: “add this to…” - Telegram from your phone (Shown: A screenshot from the tab, filed into the right note.) - The tab. Two minutes before a call: “What did we leave open with Northwind last time?” → Three open points and who owes what, while the call connects. - Attach a file. A client sends a 30-page statement of work: “Summarise this SOW and add its deadlines to Globex.” → A half-page summary in Globex’s note and four dates in the timeline. - Telegram. Leaving a client workshop: “Voice note: “Marc wants the board summary in Spanish too.”” → Filed under Acme as a task, before you reach the station. - Search ⌘F. You remember a phrase, not the meeting: ““regional churn”” → The workshop note and the line it’s on, in a second. ### Schedule routines: your AI sends you a weekly summary of every client on its own Tell your agent, once, what you want to know and when: “every Friday at four, go through every client and tell me what’s due, what’s waiting and what I haven’t invoiced”. At that time it starts the conversation on its own. When there’s nothing to say, it stays quiet. Routines run while Glowstone is open on your Mac. - Set in plain words, no forms - It starts the conversation, you don’t - Silent when there’s nothing to report (Shown: Friday at 16:00, the wrap-up arrives on its own.) Example. Friday, and five clients each have something pending. You ask: “Once: “every Friday at four, tell me what’s due, waiting and not invoiced across clients.”” Glowstone: Writes the routine into your workspace; Every Friday at 16:00, reads every client’s notes; Opens a conversation with what needs you. Result: Nothing slips into Monday: the overdue reply to Sophie and the unsent invoice show up before the weekend. ### Connect Notion, Slack, Stripe and more, so your AI can read and update them for you Most tools you work in today can be connected to an AI agent, through what are called MCP connectors. Tell Glowstone which ones you use and your agent sets them up with you, in the conversation: it looks for the official connector, adds it, and walks you through signing in. No settings screen, no developer needed. From then on it can read them and, where the tool allows it, act in them: create and update tickets, write pages, draft emails, check a customer’s invoices. Every action shows up in the chat, and it only reaches what your own account can. - Connected in the conversation, guided step by step - Reads and, where allowed, writes in your tools - Every action visible in the chat - **Notion:** Read and write client pages, task lists and meeting notes - **Slack:** Read channels and post updates (with a token) - **Stripe:** Check invoices and what each client has paid - **Jira & Confluence:** Read and create issues and pages, in clients that use Atlassian - **Linear:** Read and create issues for client projects - **Calendar & email:** Read your day and draft emails, through community connectors - **Google Drive:** Find and read client documents - **Asana:** Read and update client projects and tasks (Shown: The steering committee turned into Notion tasks, one by one.) Example. Every client lives in a different tool: one in Notion, one in Jira, one in email. You ask: “What’s open for Globex in their Jira, and what did we promise them last week?” Glowstone: Reads Globex’s Jira issues through the Atlassian connector; Reads last week’s meeting note; Matches promises against open issues. Result: One view across their tool and your notes: two promises without an issue yet, which you create on the spot. Also connectable (official or community connectors): monday.com, Xero, Canva, Amplitude, Zapier. Dictation: speak instead of typing, to your agent or in any app, about three times faster. Free, with Whisper running on the Mac. ### A day with Glowstone - 08:45: The morning routine lists today’s client calls and what’s due. - 10:00: Steering committee with Acme. Recorded from the tab, a question answered mid-call. - 11:05: Process: decisions and follow-ups filed into Acme’s notes. - 13:30: The proposal for Globex, dictated instead of typed. - 16:00: Before a Northwind call: “what did we leave open last time?” - Fri 16:00: The weekly wrap-up across every client, and what isn’t invoiced. ### Things to ask - “What’s open with each client, and who is it waiting on?” - “What did we agree with Acme about scope in September?” - “Prepare me for the Northwind call in ten minutes.” - “Draft the steering committee summary from today’s meeting.” - “Which deliverables are due this week?” - “What hours or work haven’t I invoiced yet?” ### Questions - **Can I keep clients separate?** Yes. Each client can have its own notes inside one workspace, or you can use separate spaces: each space is its own folder with its own conversations, and the agent in one never sees the other. - **Does the client see a bot in the call?** No. Glowstone listens on your Mac: your microphone for you, the system audio for everyone else. It works with any call app because it never joins the call. Recording people is regulated in many places, so tell them first. - **Where does client information go?** Your notes are files on your Mac and Glowstone has no server that receives them. When you ask something, the agent you chose (for example Claude Code, from Anthropic) reads the notes it needs to answer, under your own account. Transcription and dictation run on your Mac. - **Is it free?** Glowstone is free, and so is transcription and dictation. The agent you chat with uses your own plan with its provider, as it does today. ## The AI second brain for product managers https://glowstone.ai/for/product-managers Every decision, with its why. Glowstone is like an assistant who was in every meeting and read every note about your products: the decisions, the reasons, the customer promises and who owns what. It thinks with the judgment of your best senior PM and answers from your own records, so “why did we decide that?” always has an answer. ### The problem - **Decisions without a trail.** It was agreed in a meeting three weeks ago. Nobody remembers why, and the doc that should say it doesn’t. - **Promises to customers, scattered.** A date said on a sales call, a feature mentioned in a renewal. Then planning happens without it. - **Status updates eat your week.** Gathering what changed across teams, tickets and threads is a job in itself, every single week. ### Keep every product decision on record and ask your AI anything about it Each product, person, customer and decision gets a note, written and linked by your agent as you work. Specs and tickets stay where they are; Glowstone keeps the part that usually lives nowhere: why things were decided, who asked for what, what was promised. Then you ask it, like you’d ask the one teammate who was in every meeting: what’s blocking a release, what a customer was told, what changed since last planning. - Decisions saved with the why and the date - Customer promises linked to the product they affect - With Linear connected, tickets join the answer (Shown: A decision note inside Glowstone, then a question answered from the product’s history.) Example. Leadership asks why the dashboard redesign slipped two weeks. The decision was made in a meeting you half remember. You ask: “Why did we move the Atlas redesign, and who agreed?” Glowstone: Finds the decision note from Oct 5 sprint planning; Reads the Northwind note behind it; Answers with the date, the people and the reason. Result: “Oct 5, with Maya and Leo: CSV export was promised to Northwind before their renewal.” Copy, paste, done. ### Record planning and customer calls and ask questions live, with the meeting and your product history as context Record sprint planning, a design review or a customer call from the tab. No bot joins; Glowstone transcribes on your Mac and splits what you said from what the others said. Mid-meeting, ask what the room can’t answer: what a customer was told, what was decided last quarter, what Linear says. Afterwards, Process saves the decisions and updates the products and people they touch. - Ask mid-call: promises, past decisions, tickets - Decisions filed with the date and the reason - Ask weeks later what someone said, exactly (Shown: Sprint planning: the answer comes from a customer promise and a Linear issue.) Example. In planning, the team proposes dropping CSV export to protect the redesign. You ask: “Did we promise Northwind anything before their renewal?” Glowstone: Reads the call transcript so far; Opens Northwind’s note and your Sep 9 call; Checks Linear for open export issues. Result: You catch the promise before the team drops it, and leave with a plan that keeps the customer. ### Use it all day: capture dashboards, jot ideas and ask from any app Most of the value comes from small moments, dozens of times a day. Glowstone hangs from the top of your screen in every app: bring the pointer up, ask, jot something down or attach what’s on screen, and get back to work. Away from the desk, a voice note to your own Glowstone bot on Telegram does the same. Everything lands in the same memory. - The tab, over any app - Screenshots it understands: “add this to…” - Telegram from your phone (Shown: A screenshot from the tab, filed into the right note.) - The tab. An idea in the middle of a review: “Note: export should remember the last date range.” → Added to Atlas’s ideas, linked to CSV export. - Screenshot gesture. An error in staging: “↓ + → and “what does this mean?”” → A new chat with the screenshot attached and the likely cause. - Attach a file. Research interviews arrive as a PDF: “Pull the top three pains and link them to Beacon.” → Three insights in Beacon, each with its quote. - Telegram. Walking out of a customer visit: “Voice note: “Northwind wants SSO before anything else.”” → Filed in Northwind and flagged for planning. ### Schedule routines: your AI sends you the Monday product status on its own Ask once: “every Monday at nine, tell me what changed across my products last week, what’s blocked and which customer promises are coming due”. It starts that conversation on its own, every week. Silent when there’s nothing new. Routines run while Glowstone is open on your Mac. - Set in plain words - It starts the conversation, you don’t - With connectors, it reads tickets and calendar too (Shown: Monday at 9:00, the product status arrives on its own.) Example. Every Monday you rebuild the status across two products from tickets, threads and notes. You ask: “Once: “every Monday at nine, tell me what changed, what’s blocked and which promises are due.”” Glowstone: Reads last week’s notes and meetings; With Linear connected, reads the tickets; Opens a conversation with the summary. Result: Monday’s status is waiting for you, and the draft update for leadership is one question away. ### Connect Linear, Jira, Sentry and more, so your AI can read and create tickets for you Most tools you work in today can be connected to an AI agent, through what are called MCP connectors. Tell Glowstone which ones you use and your agent sets them up with you, in the conversation: it looks for the official connector, adds it, and walks you through signing in. No settings screen, no developer needed. From then on it can read them and, where the tool allows it, act in them: create and update tickets, write pages, draft emails, check a customer’s invoices. Every action shows up in the chat, and it only reaches what your own account can. - Connected in the conversation, guided step by step - Reads and, where allowed, writes in your tools - Every action visible in the chat - **Linear:** Read, create and update issues, projects and cycles - **Jira & Confluence:** Read and write issues and sprints, read and edit specs - **Notion:** Read and write specs, roadmaps and docs - **Sentry:** See what’s breaking after a release - **GitHub:** Read pull requests and issues (with a token) - **Slack:** Read threads and post updates (with a token) - **Amplitude:** Read funnels, charts and user behaviour - **PostHog:** Product analytics, feature flags and session data (Shown: Planning turned into Jira tickets, assigned and linked to the decision.) Example. After a release, support pings you about errors and you don’t know if they’re new. You ask: “Is anything new breaking since Tuesday’s Beacon release, and is there a ticket for it?” Glowstone: Reads Sentry issues since the release; Searches Linear for matching tickets; Reads the release note in your vault. Result: One new crash on Android onboarding with no ticket yet. It drafts the ticket with the stack trace for you to approve. Also connectable (official or community connectors): Mixpanel, Google Analytics, Figma, Asana, monday.com, Google Search Console, Intercom, Zapier. Dictation: speak instead of typing, to your agent or in any app, about three times faster. Free, with Whisper running on the Mac. ### A day with Glowstone - 09:00: Monday status across both products, from a routine. - 11:00: Sprint planning recorded. A customer promise checked mid-call. - 11:40: Process: the decision saved with its reason; Atlas and the team updated. - 14:00: A spec’s first draft, spoken instead of typed. - 16:30: “What did Northwind ask for in the last three calls?” - 18:00: A note from the tab before closing the laptop. ### Things to ask - “Why did we decide to move the redesign?” - “What have we promised customers this quarter, and when?” - “What changed on Atlas since last planning?” - “Summarise this week’s user tests for Beacon.” - “Draft the stakeholder update from this week’s meetings.” - “Who owns each action item from today’s planning?” ### Questions - **Does this replace Linear or Jira?** No. Tickets stay in your tracker. Glowstone keeps what usually lives nowhere: why things were decided, what was promised to whom, what was said in the meeting. With the Linear connector, your agent reads tickets when an answer needs them. - **Can my team use it?** Glowstone is a personal workspace on your Mac today: there’s no shared server. You can share notes like any files, but it’s built for your own memory, not a team wiki. - **Does a bot join our calls?** No. Glowstone listens on your Mac, through your microphone and the system audio, and works with any call app. Tell people you’re recording. - **Is it free?** Yes. Glowstone, transcription and dictation are free. The agent uses your own plan with its provider. - **Can I handle customer support with it too?** Yes, and it pairs well with product work: every customer question lands in the same memory as your decisions and roadmap. Keep policies and past rulings as notes, connect your help desk, and draft replies in each customer’s language. See Glowstone for customer support. ## The AI second brain for real estate agents https://glowstone.ai/for/real-estate-agents Every buyer, every listing, matched. Glowstone is like an assistant who remembers every buyer, seller, listing and viewing you’ve ever had: budgets, must-haves, offers and conditions. It thinks with the judgment of your most experienced colleague and answers from your own records, so the right buyer is one question away. ### The problem - **Buyers you can’t keep in your head.** Budget, neighbourhood, must-haves, what they said at the last viewing. Twenty of them, and a new listing that suits three. - **Notes in the car, after the viewing.** The details you wanted to write down are gone by the time you park, or end up in voice memos nobody listens to again. - **Offers and conditions, everywhere.** One offer by phone, a counter by email, a condition in a message. The full picture lives in no single place. ### Keep every listing, buyer and seller on record and ask your AI anything about them Each listing, buyer, seller and viewing gets a note your agent writes and links as you work: what the buyer needs, what the seller will accept, what was said at each visit. Then ask in plain words: which buyers would want a new listing, what the best offer on a flat is and with which conditions, who you haven’t called back this week. - One note per property, buyer, seller and viewing - Budgets, must-haves and conditions, all linked - Answers that say where they came from (Shown: A listing inside Glowstone, then the buyers that fit it, with their reasons.) Example. A new three-bedroom flat with a terrace comes in at €520k. You have twenty active buyers. You ask: “Which buyers would want Gràcia 14, and what did each say they need?” Glowstone: Reads the listing note; Goes through every buyer’s budget and must-haves; Ranks the ones that fit and says why the others don’t. Result: Two calls to make, with what to tell each, before any other agent has seen the listing. ### Record calls with buyers and sellers and ask questions live, with the call and your client history as context Record a call with a seller or a buyer from the tab. No bot joins; Glowstone transcribes on your Mac and keeps who said what apart. When they ask something you’d have to check, ask the tab mid-call: the best offer so far, what the other buyer conditioned, what the seller said last week. Afterwards, Process updates the listing, the people and the follow-ups. - Every offer and condition, recalled mid-call - No bot, nothing for the client to accept - Listings and people updated after the call (Shown: A seller call: the best offer and its conditions, recalled in seconds.) Example. The seller calls and asks where the offers stand. They change every week. You ask: “What’s the best offer on Rosselló 203 and its conditions?” Glowstone: Reads the call transcript so far; Opens the listing and each buyer’s note; Compares amounts, conditions and timing. Result: You answer the seller precisely, including which offer can actually sign before October. ### Use it all day: capture listings, jot notes and ask from any app Most of the value comes from small moments, dozens of times a day. Glowstone hangs from the top of your screen in every app: bring the pointer up, ask, jot something down or attach what’s on screen, and get back to work. Away from the desk, a voice note to your own Glowstone bot on Telegram does the same. Everything lands in the same memory. - The tab, over any app - Screenshots it understands: “add this to…” - Telegram from your phone (Shown: A screenshot from the tab, filed into the right note.) - Telegram. Right after a viewing, in the street: “Voice note: “The Hills loved the terrace, worried about noise.”” → Filed in the viewing and the Hills’ note. - The tab. A seller calls unexpectedly: “What did I last tell Marc about the floor price?” → The date and your exact words, before you pick up. - Attach a file. A buyer sends a mortgage pre-approval: “Add this to Anna’s note and tell me her real budget.” → Pre-approval saved, budget updated to €510k. - Search ⌘F. Who mentioned a garden?: ““garden”” → Two buyers and the viewings where they said it. ### Schedule routines: your AI tells you every morning who needs a follow-up Ask once: “every morning at nine, tell me which buyers and sellers I haven’t contacted in a week, and which offers expire soon”. It starts that conversation on its own. Silent when everyone is up to date. Routines run while Glowstone is open on your Mac. - Follow-ups that don’t depend on memory - Offers and deadlines flagged before they lapse - Silent when there’s nothing to do (Shown: 9:00, the day’s follow-ups arrive on their own.) Example. With dozens of contacts, some buyers go a week without hearing from you. You ask: “Once: “every morning at nine, tell me who I haven’t contacted in a week and which offers expire soon.”” Glowstone: Reads every buyer, seller and offer; Checks dates against today; Opens a conversation with the list. Result: No buyer goes cold and no offer expires unnoticed. ### Connect your calendar, email and CRM, so your AI can book viewings and draft messages for you Most tools you work in today can be connected to an AI agent, through what are called MCP connectors. Tell Glowstone which ones you use and your agent sets them up with you, in the conversation: it looks for the official connector, adds it, and walks you through signing in. No settings screen, no developer needed. From then on it can read them and, where the tool allows it, act in them: create and update tickets, write pages, draft emails, check a customer’s invoices. Every action shows up in the chat, and it only reaches what your own account can. - Connected in the conversation, guided step by step - Reads and, where allowed, writes in your tools - Every action visible in the chat - **Calendar:** Book and move viewings, through a community connector - **Email:** Draft confirmations and follow-ups, through a community connector - **HubSpot:** Read contacts, deals and their history (read-only connector) - **Salesforce:** Read and update records (official connector in pilot) - **Notion:** Keep listings and buyer pages your team shares - **Drive & documents:** Find and read listing documents, through community connectors - **monday.com:** Track listings and deals on your boards - **Xero:** Invoices and commissions (Shown: Two viewings booked and two confirmations drafted, from one sentence.) Example. Your agency keeps contacts in HubSpot, but your notes on what each buyer said live with you. You ask: “Which buyers in HubSpot asked about Gràcia, and what did they tell me at viewings?” Glowstone: Reads contacts and deals through the HubSpot connector; Reads each buyer’s viewing notes in your vault; Joins both into one list. Result: The CRM’s data and your own judgment in one answer, without copying anything by hand. Also connectable (official or community connectors): Google Drive, Canva, Asana, Zapier. Dictation: speak instead of typing, to your agent or in any app, about three times faster. Free, with Whisper running on the Mac. ### A day with Glowstone - 09:00: Morning follow-ups: who’s waiting, which offers expire. - 11:00: A new listing comes in: “which buyers would want this?” - 12:30: Viewing notes dictated in the car park. - 17:20: Seller call recorded. The best offer recalled mid-call. - 17:45: Process: the listing and both buyers updated. - 20:00: A voice note to Telegram after a late viewing. ### Things to ask - “Which buyers fit this new listing?” - “What offers do we have on Rosselló 203, and with which conditions?” - “Who haven’t I called back this week?” - “What did the Hills say they disliked at the last viewing?” - “Draft the listing description from my viewing notes.” - “Which offers expire this week?” ### Questions - **Does it replace my CRM?** Not necessarily. Glowstone is where the context lives: what each person said and wants, every call and viewing. If your CRM has an MCP connector, your agent can read it too. - **Can I use it from my phone?** Yes, through Telegram: write or send a voice note to your own Glowstone bot and it’s answered and filed by your Mac. It answers while Glowstone is open on your Mac. - **Where does client information go?** Your notes are files on your Mac; Glowstone has no server. The agent you chose reads what it needs to answer, under your account. Transcription and dictation run on your Mac. - **Is it free?** Yes. Glowstone, transcription and dictation are free. The agent uses your own plan with its provider. ## The AI second brain for founders https://glowstone.ai/for/founders The whole company, in one memory. Glowstone is like a chief of staff who has read everything about your company: every investor conversation, candidate, customer and decision. It thinks with the judgment of a seasoned operator and answers from your own records, so the company’s context lives somewhere other than your head. ### The problem - **Ten threads, one head.** A fundraise, three hires, the biggest customer’s renewal and a product launch. The details of each are spread across inboxes, chats and memory. - **What did we tell them?.** Numbers shared with an investor, terms discussed with a candidate. Saying something different next time costs trust. - **Updates nobody has time to write.** The investor update, the board memo, the all-hands notes. Assembling them means rereading a month of conversations. ### Keep investors, hires and customers on record and ask your AI anything about them Each investor, candidate, customer and decision gets a note your agent writes and links as you work: what was said, what was shared, what’s pending and with whom. Then ask: where each investor in the round stands, what you told a fund about burn last month, which candidates are still in the process and what they asked for. - Investors, hiring, customers and decisions, linked - What you shared, with whom and when - Answers that cite the conversation they came from (Shown: The round inside Glowstone, then where every investor stands.) Example. Mid-fundraise, with nine investors at different stages and a board call tomorrow. You ask: “Where does each investor in the round stand, and what do they still need from us?” Glowstone: Reads the round note and each investor’s note; Pulls what each one asked for and when; Adds up what’s committed and what’s left. Result: A clean pipeline of the round, with the two things you owe someone this week. ### Record investor, customer and candidate calls and ask questions live, with the call and your company history as context Record from the tab: investor meetings, candidate interviews, customer calls. No bot joins; Glowstone transcribes on your Mac. Mid-call, check what you said before: the numbers you shared, the terms you discussed, what the customer was promised. Afterwards, Process files it all where it belongs. - What you told each investor, recalled mid-call - Candidate interviews kept and comparable - Customer promises never lost (Shown: An investor call: last month’s numbers, recalled exactly.) Example. A partner asks, mid-call, what burn you shared last month. You ask: “What burn and runway did we share with Northbeam in September?” Glowstone: Reads the call so far; Opens Northbeam’s note and the Sep 12 meeting; Compares it with September’s actual numbers. Result: You give the same figure you gave before, and explain the difference with confidence. ### Use it all day: capture dashboards, jot thoughts and ask from any app Most of the value comes from small moments, dozens of times a day. Glowstone hangs from the top of your screen in every app: bring the pointer up, ask, jot something down or attach what’s on screen, and get back to work. Away from the desk, a voice note to your own Glowstone bot on Telegram does the same. Everything lands in the same memory. - The tab, over any app - Screenshots it understands: “add this to…” - Telegram from your phone (Shown: A screenshot from the tab, filed into the right note.) - The tab. Between two investor calls: “What did Fieldstone ask for last time?” → September’s numbers, promised by Friday. You send them now. - Attach a file. A term sheet arrives: “Summarise the key terms and compare with Northbeam’s.” → A side-by-side of valuation, pro rata and board seat. - Telegram. After a coffee with a candidate: “Voice note: “Irene can start in January, wants equity over salary.”” → Filed in the design lead hiring note. - Search ⌘F. Where did we decide to raise $2M?: ““raise $2M”” → The decision note and the meeting it came from. ### Schedule routines: your AI sends you a weekly company review on its own Ask once: “every Friday at five, summarise the week: the round, hiring, our top customers and anything I promised someone”. It starts that conversation on its own, and can draft the investor update from it. Silent when there’s nothing new. Routines run while Glowstone is open on your Mac. - The week across the company, in one message - Promises you made, surfaced before they’re due - The first draft of the update, ready to edit (Shown: Friday at 17:00, the company review arrives on its own.) Example. The week was a blur of calls across the round, hiring and customers. You ask: “Once: “every Friday at five, review the round, hiring, top customers and anything I promised.”” Glowstone: Reads the week’s meetings and notes; Lists what moved and what’s due; Offers to draft the investor update. Result: You close the week knowing where everything stands, with a draft ready to edit. ### Connect Stripe, Notion, Linear and more, so your AI can pull numbers and update documents for you Most tools you work in today can be connected to an AI agent, through what are called MCP connectors. Tell Glowstone which ones you use and your agent sets them up with you, in the conversation: it looks for the official connector, adds it, and walks you through signing in. No settings screen, no developer needed. From then on it can read them and, where the tool allows it, act in them: create and update tickets, write pages, draft emails, check a customer’s invoices. Every action shows up in the chat, and it only reaches what your own account can. - Connected in the conversation, guided step by step - Reads and, where allowed, writes in your tools - Every action visible in the chat - **Stripe:** Revenue, customers, subscriptions and invoices - **Notion:** Read and write the investor update, board docs and wiki - **Linear or Jira:** See what the team is shipping and what’s stuck - **HubSpot:** Read the sales pipeline (read-only connector) - **GitHub:** Read pull requests and releases (with a token) - **Slack:** Read threads and post updates (with a token) - **Xero:** Accounting: invoices, payments and cash - **Google Analytics:** Traffic, sign-ups and conversions (Shown: Stripe’s numbers written into the Notion draft, section by section.) Example. The board asks how engineering is going and you haven’t looked at the tracker in a week. You ask: “What did the team ship this month, and what’s stuck?” Glowstone: Reads closed and open issues through the Linear connector; Reads the product notes in your vault; Groups them by project. Result: A three-line answer for the board, and the one stuck project you should ask about. Also connectable (official or community connectors): Amplitude, Google Search Console, Intercom, Figma, Zapier. Dictation: speak instead of typing, to your agent or in any app, about three times faster. Free, with Whisper running on the Mac. ### A day with Glowstone - 08:30: The morning brief: calls today and what each one needs. - 10:00: Candidate interview recorded, compared later with the others. - 15:00: Investor call. Last month’s numbers checked mid-call. - 15:50: Process: the round updated, a promise added to the list. - 18:00: The investor update, dictated. - Fri 17:00: The company review, with a draft update. ### Things to ask - “Where does each investor in the round stand?” - “What numbers did we share with Northbeam, and when?” - “Compare the design lead candidates after the final round.” - “What have I promised customers this quarter?” - “Draft this month’s investor update.” - “What did we decide about pricing, and why?” ### Questions - **Is it safe to put company information in it?** Your notes are files on your Mac; Glowstone has no server and can’t read them. The agent you choose (Claude Code, Codex…) reads what it needs to answer, under your own account and its provider’s terms. Choosing the agent is choosing who sees it. - **Can I keep personal life separate?** Yes. Use spaces: each is its own folder with its own conversations, and the agent in one never sees the other. - **Does a bot join investor calls?** No. Glowstone listens on your Mac and works with any call app. Tell people you’re recording. - **Is it free?** Yes. Glowstone, transcription and dictation are free. The agent uses your own plan with its provider. - **Can I run customer support from it?** Yes. Early on, the founder is often the support team. Keep your policies and the exceptions you make in Glowstone, connect Stripe or your help desk, and ask it to draft replies in each customer’s language. See Glowstone for customer support for the full picture. ## The AI second brain for salespeople https://glowstone.ai/for/sales Every account, every promise, kept. Glowstone is like an assistant who has listened to every call and read every note about your accounts: needs, objections, who decides and what you promised. It thinks with the judgment of your best account executive and answers from your own records, so you never walk into a call cold. ### The problem - **Calls without context.** You last spoke three weeks ago. What did they push back on? Who else was in the room? The CRM has a stage and a date, not the conversation. - **Follow-ups written at 9 pm.** Every call deserves a recap, and every recap is ten minutes of typing you don’t have between calls. - **Promises that slip.** “I’ll check with product” was said on a call and written nowhere. The customer remembers it at renewal. ### Keep every account and call on record and ask your AI anything about them Each account, contact and call gets a note your agent writes and links as you work: needs, objections, who decides, what was promised. With your CRM or Stripe connected, the agent reads those too. Then ask: which deals close this month and what’s blocking each, what a customer pays, what you promised them on the last call. - Accounts, contacts, calls and promises, linked - Objections and decision makers remembered - CRM and billing data with connectors (Shown: A deal inside Glowstone, then the month’s closes and their blockers.) Example. Forecast call with your manager in ten minutes. You ask: “Which deals close this month, and what’s blocking each?” Glowstone: Reads every deal and account note; Finds the close dates and open blockers; Notes who decides and where they stand. Result: A forecast you can defend, deal by deal, with the next step for each. ### Record sales calls and ask questions live, with the call and the account’s history as context Record any call from the tab. No bot joins, so the customer sees nothing different; Glowstone transcribes on your Mac. When they ask what they’d pay or what was agreed, ask the tab: it reads your notes and, if you connected them, your billing or CRM. Afterwards, Process files the call and adds the follow-ups. - Pricing and history, checked mid-call - Billing and CRM data through connectors - Every call filed, every follow-up listed (Shown: A renewal call: the floor price and the account’s status, from a note and Stripe.) Example. On a renewal, the customer asks for more seats at their old price. You ask: “Can I keep Northwind at their price for 150 seats? Are they up to date?” Glowstone: Reads your pricing note; Checks the account in Stripe through your connector; Works out your floor price. Result: You give a confident number on the call, inside your discount limit. ### Use it all day: capture profiles and pages, jot notes and ask from any app Most of the value comes from small moments, dozens of times a day. Glowstone hangs from the top of your screen in every app: bring the pointer up, ask, jot something down or attach what’s on screen, and get back to work. Away from the desk, a voice note to your own Glowstone bot on Telegram does the same. Everything lands in the same memory. - The tab, over any app - Screenshots it understands: “add this to…” - Telegram from your phone (Shown: A screenshot from the tab, filed into the right note.) - The tab. Thirty seconds before a call: “Who decides at Acme and what do they care about?” → Hannah decides; she cares about rollout time. Lead with that. - Screenshot gesture. A LinkedIn profile of a new contact: “↓ + → and “add as contact for Globex”” → A new person note linked to the account. - Telegram. In the taxi after a demo: “Voice note: “Tom wants a pilot for 20 seats first.”” → Filed in Northwind with a follow-up for tomorrow. - Search ⌘F. Which customer mentioned a competitor?: ““competitor”” → Two accounts and the calls where it came up. ### Schedule routines: your AI sends you a pipeline brief every morning Ask once: “every weekday at 8:30, tell me today’s calls, what each account needs and which follow-ups are overdue”. It starts that conversation on its own, before your first call. Silent when there’s nothing to do. Routines run while Glowstone is open on your Mac. - Today’s calls with their context - Overdue follow-ups, before they cost a deal - With your calendar connected, it knows who’s coming (Shown: 8:30, the day’s pipeline arrives on its own.) Example. Promised follow-ups get lost between calls. You ask: “Once: “every weekday at 8:30, tell me today’s calls, what each needs and what’s overdue.”” Glowstone: Reads your calendar if connected, and your accounts; Finds follow-ups past their date; Opens a conversation with the day. Result: You start the day knowing every call and every promise due. ### Connect HubSpot, Salesforce, Stripe and more, so your AI can read your pipeline and accounts for you Most tools you work in today can be connected to an AI agent, through what are called MCP connectors. Tell Glowstone which ones you use and your agent sets them up with you, in the conversation: it looks for the official connector, adds it, and walks you through signing in. No settings screen, no developer needed. From then on it can read them and, where the tool allows it, act in them: create and update tickets, write pages, draft emails, check a customer’s invoices. Every action shows up in the chat, and it only reaches what your own account can. - Connected in the conversation, guided step by step - Reads and, where allowed, writes in your tools - Every action visible in the chat - **HubSpot:** Read contacts, companies, deals and tickets (read-only connector) - **Salesforce:** Read and update records (official connector in pilot) - **Stripe:** What a customer pays, and whether they’re up to date - **Email:** Draft follow-ups, through a community connector - **Calendar:** See who you’re meeting, through a community connector - **Slack:** Read deal channels and post updates (with a token) - **monday.com:** Pipelines and accounts on your boards - **Notion:** Account plans, battle cards and playbooks (Shown: HubSpot’s pipeline joined with your own call notes.) Example. A customer asks for a discount and you need their history fast. You ask: “How much has Globex paid us this year, and are they behind on anything?” Glowstone: Reads Globex’s invoices through the Stripe connector; Reads the account note in your vault; Checks for open disputes or overdue invoices. Result: Their total for the year, one invoice ten days late, and the discount you gave them in March. Enough to decide on the call. Also connectable (official or community connectors): Intercom, Google Drive, Canva, Mixpanel, Xero, Zapier. Dictation: speak instead of typing, to your agent or in any app, about three times faster. Free, with Whisper running on the Mac. ### A day with Glowstone - 08:30: The pipeline brief: calls today and overdue follow-ups. - 10:00: Discovery call recorded. A question about pricing answered mid-call. - 10:40: Process: the account and the deal updated, follow-ups listed. - 10:45: The recap, dictated and sent before the next call. - 16:00: Renewal call: what they pay, checked in Stripe. - 18:00: “What did I promise anyone today?” ### Things to ask - “Which deals close this month, and what’s blocking each?” - “What did Northwind object to on the last call?” - “Prepare me for the Acme call at 10.” - “What have I promised customers and not done yet?” - “Draft the recap from this call.” - “Who decides at Globex, and have I met them?” ### Questions - **Does it replace my CRM?** No. The CRM keeps the stage and the numbers; Glowstone keeps the conversation: what was said, objected and promised. If your CRM offers an MCP connector, your agent can read it too. - **Does the customer see a bot?** No. Glowstone listens on your Mac and works with any call app. Tell people you’re recording. - **Where does customer information go?** Your notes are files on your Mac; Glowstone has no server. The agent you choose reads what it needs to answer, under your account. Transcription and dictation run on your Mac. - **Is it free?** Yes. Glowstone, transcription and dictation are free. The agent uses your own plan with its provider. ## The AI second brain for students https://glowstone.ai/for/students Every class, remembered. Glowstone is like a study partner who attended every lecture and read every note you took: every subject, reading, deadline and what the professor said would be on the exam. It thinks with the judgment of the best student in your class and answers from your own notes, so revision starts from what you actually covered. ### The problem - **Notes that don’t add up.** Half in a notebook, half in a doc, some on your phone. When the exam comes, you rebuild the course from fragments. - **Lectures you can’t replay.** The professor said something important about the exam, and you were still copying the previous slide. - **Deadlines in five places.** The syllabus, an email, a group chat, the course page. Something always surprises you on a Sunday night. ### Keep every subject, class and reading on record and ask your AI anything about them Each subject gets its notes: what was covered in every lecture, the readings, your doubts, the deadlines and what the professor stressed. Your agent writes and links them as you study, so the course organises itself. Then ask it like a classmate who never missed a class: what will be on the midterm, which lectures cover a topic, what you still don’t understand. - One note per subject, lecture and reading - Exam hints and deadlines saved with the date - Answers that cite the lecture they came from (Shown: A subject’s notes inside Glowstone, then a question answered from your own lectures.) Example. It’s the night before the Statistics exam. You ask: “Make me 15 practice questions from my Statistics notes, hardest topics first.” Glowstone: Reads every Statistics lecture and reading; Finds the topics you marked as unclear; Writes the questions, with answers at the end. Result: A practice test built from what was actually taught in your class, not from a generic textbook. ### Record lectures and group meetings and ask questions live, with the class and your notes as context Start recording from the tab when the class starts. In an online lecture Glowstone hears your microphone and the lecturer separately; in a room, everything comes through the microphone as one track. It’s transcribed on your Mac, so nothing goes to a server. Lost? Ask the tab during the class, with what’s been said so far and your notes behind the answer. Afterwards, Process turns the lecture into notes for that subject. Check your university’s rules and ask the lecturer before recording. - The full lecture, transcribed on your Mac - Ask mid-class without interrupting anyone - Lecture notes filed into the right subject (Shown: A lecture: the professor explains, you ask the tab, the answer links to last week’s class.) Example. Your group project meeting ends and nobody wrote down who does what. You ask: “Who agreed to do what in today’s group meeting, and by when?” Glowstone: Reads the meeting transcript; Pulls out each commitment and the person; Adds the dates to the project’s deadlines. Result: A clear list for the group chat: Ana the survey by Thursday, you the analysis by Sunday, Marc the slides. ### Use it all day: capture slides, jot ideas and ask from any app Most of the value comes from small moments, dozens of times a day. Glowstone hangs from the top of your screen in every app: bring the pointer up, ask, jot something down or attach what’s on screen, and get back to work. Away from the desk, a voice note to your own Glowstone bot on Telegram does the same. Everything lands in the same memory. - The tab, over any app - Screenshots it understands: “add this to…” - Telegram from your phone (Shown: A screenshot from the tab, filed into the right note.) - The tab. Reading a dense paper: “Explain this paragraph simply, and save it to Marketing.” → A plain explanation, filed next to the reading. - Attach a file. The syllabus arrives as a PDF: “Add every deadline and exam date from this to my term.” → Nine dates in This term.md, ready for your calendar. - Telegram. An idea on the bus: “Voice note: “For the essay, compare Spain and Portugal’s minimum wage.”” → Filed in the Microeconomics essay note. - Search ⌘F. Where did we see elasticity?: ““elasticity”” → Three lectures and the reading, in a second. ### Schedule routines: your AI sends you every Sunday what’s due this week Ask once: “every Sunday at seven, tell me what’s due this week, what exams are coming and what I haven’t reviewed”. It starts that conversation on its own, every week. Silent when there’s nothing due. Routines run while Glowstone is open on your Mac. - A weekly plan you don’t have to build - Exams flagged weeks before, not days - Silent when there’s nothing due (Shown: Sunday at 19:00, the week’s plan arrives on its own.) Example. Exams are three weeks away and you don’t know where to start. You ask: “Once: “every evening at eight during exams, give me tomorrow’s study plan.”” Glowstone: Reads every subject and what you’ve reviewed; Weighs the exam dates; Opens a short plan each evening. Result: A realistic plan every night, adjusted to what you actually did that day. ### Connect your calendar, Notion, Google Drive and more, so your AI can organise your studies for you Most tools you work in today can be connected to an AI agent, through what are called MCP connectors. Tell Glowstone which ones you use and your agent sets them up with you, in the conversation: it looks for the official connector, adds it, and walks you through signing in. No settings screen, no developer needed. From then on it can read them and, where the tool allows it, act in them: create and update tickets, write pages, draft emails, check a customer’s invoices. Every action shows up in the chat, and it only reaches what your own account can. - Connected in the conversation, guided step by step - Reads and, where allowed, writes in your tools - Every action visible in the chat - **Notion:** Read and write your study pages and databases - **Google Drive:** Find and read course documents - **Calendar:** Add deadlines and exams, through a community connector - **Email:** Find what your professors sent, through a community connector - **GitHub:** Read your coding assignments and repos (with a token) - **Asana:** Plan group projects with tasks and owners - **Slack:** Read your class or group channels (with a token) - **Zapier:** Connect thousands of other apps, including many course tools (Shown: Deadlines from your notes into your calendar.) Example. Your group keeps its project in Notion and nobody updates it. You ask: “Update our Notion project page with what we decided today, and who does what.” Glowstone: Reads today’s group meeting; Finds the project page through the Notion connector; Writes the decisions and the owners. Result: The shared page is up to date without anyone typing it. Also connectable (official or community connectors): Linear, Figma, Canva, monday.com. Dictation: speak instead of typing, to your agent or in any app, about three times faster. Free, with Whisper running on the Mac. ### A day with Glowstone - 08:30: A look at today: classes, what’s due, what to review. - 10:00: Lecture recorded. A question asked mid-class, answered from your notes. - 11:30: Process: the lecture filed into its subject. - 16:00: A slide captured from the tab and added to your notes. - 19:00: Group meeting: who does what, written down for you. - Sun 19:00: The weekly study plan arrives on its own. ### Things to ask - “What will be on the midterm, according to the professor?” - “Make me a practice test from my Statistics notes.” - “Explain consumer surplus with the example from class.” - “What’s due this week, across all subjects?” - “What did I mark as unclear and haven’t reviewed?” - “Summarise lecture 6 in ten lines.” ### Questions - **Is it free for students?** Glowstone is free, and so are transcription and dictation. The agent needs a plan with its provider, but OpenCode and Kilo Code can run on free models, so you can use Glowstone without paying anything. - **Can I record lectures?** Technically yes: online classes and in-room lectures. Whether you may depends on your university and the lecturer, so check the rules and ask first. - **Will it write my essays?** It works with your own notes and what you ask it. Use it to understand, organise and revise; follow your university’s rules on using AI in assessed work. - **Does it work offline?** Your notes, transcription and dictation work on your Mac without a connection. Answering questions needs your AI agent, which runs online. ## The AI second brain for recruiters https://glowstone.ai/for/recruiters Every candidate, remembered. Glowstone is like an assistant who sat in every interview and read every note about your candidates and roles: skills, salary expectations, what each hiring manager really wants. It thinks with the judgment of your most experienced recruiter and answers from your own records, so the right candidate is one question away. ### The problem - **Interviews you can’t compare.** Five candidates for one role, interviewed over two weeks. By Friday, they blur together. - **What the hiring manager really meant.** The job description says one thing; the intake call said another. The nuance lives in your memory. - **Candidates who go cold.** Someone great is waiting for feedback, and a week passes without a message. ### Keep every candidate, role and client on record and ask your AI anything about them Each role, candidate and hiring manager gets a note your agent writes and links as you work: skills, expectations, notice period, what was said in each interview and what the client really wants. Then ask: who are the best three for a role and why, who has done a specific thing, who is waiting on you. - One note per role, candidate and client - Interview notes saved and comparable - Answers that cite the interview they came from (Shown: A role inside Glowstone, then the shortlist ranked from your interview notes.) Example. A client asks for an update on a role you’re running with eight candidates. You ask: “Write a status update for Sophie on the backend role.” Glowstone: Reads the role and every candidate’s note; Checks interview dates and next steps; Writes a short client-ready update. Result: A clear update in Sophie’s language: who’s in final round, who dropped and why, next steps. ### Record interviews and intake calls and ask questions live, with the call and your candidate history as context Record interviews and intake calls from the tab. No bot joins; Glowstone transcribes on your Mac and keeps what you said apart from what the candidate or client said. Tell people you’re recording. Mid-call, check what you need: what another candidate said about the same topic, what the client asked for, what salary someone mentioned. Afterwards, Process files the interview into the candidate’s note. - Every interview transcribed and comparable - Client requirements checked mid-call - Candidate notes updated after every call (Shown: An intake call: the client’s must-haves, checked against your pipeline in seconds.) Example. Three final-round interviews for one role, on different days. You ask: “Compare the three finalists on what Sophie cares about.” Glowstone: Reads the three interview transcripts; Uses the role’s must-haves as criteria; Builds a side-by-side with quotes. Result: A comparison you can send the client, with each candidate’s own words as evidence. ### Use it all day: capture profiles, jot notes and ask from any app Most of the value comes from small moments, dozens of times a day. Glowstone hangs from the top of your screen in every app: bring the pointer up, ask, jot something down or attach what’s on screen, and get back to work. Away from the desk, a voice note to your own Glowstone bot on Telegram does the same. Everything lands in the same memory. - The tab, over any app - Screenshots it understands: “add this to…” - Telegram from your phone (Shown: A screenshot from the tab, filed into the right note.) - The tab. A candidate calls back unexpectedly: “What did I tell Tom about the salary range?” → Your exact words from last week, before you pick up. - Attach a file. A CV arrives by email: “Summarise this CV and tell me which open role it fits.” → A five-line summary and the best match, saved as a candidate note. - Telegram. After a coffee with a candidate: “Voice note: “Priya would consider a counter-offer, wants remote.”” → Filed in Priya’s note and flagged on the role. - Search ⌘F. Who mentioned leadership experience?: ““led a team”” → Four candidates and the interviews where they said it. ### Schedule routines: your AI tells you every morning who is waiting on feedback Ask once: “every morning at nine, tell me which candidates are waiting for feedback, today’s interviews and which offers are pending”. It starts that conversation on its own. Silent when everyone is up to date. Routines run while Glowstone is open on your Mac. - No candidate waits a week for feedback - Offers and deadlines flagged - Silent when there’s nothing to do (Shown: 9:00, the day’s pipeline arrives on its own.) Example. You run twelve roles across four clients. You ask: “Once: “every Friday at five, a status per client: roles, finalists and blockers.”” Glowstone: Reads every role and its candidates; Groups them by client; Opens one summary per client. Result: Friday client updates ready to send, one per client. ### Connect your ATS, calendar and email, so your AI can schedule interviews and draft messages for you Most tools you work in today can be connected to an AI agent, through what are called MCP connectors. Tell Glowstone which ones you use and your agent sets them up with you, in the conversation: it looks for the official connector, adds it, and walks you through signing in. No settings screen, no developer needed. From then on it can read them and, where the tool allows it, act in them: create and update tickets, write pages, draft emails, check a customer’s invoices. Every action shows up in the chat, and it only reaches what your own account can. Greenhouse, Workable, Ashby, Teamtailor and Gem have their own official connectors: your agent reads candidates, roles and feedback there, and in most of them can also move a stage or add a note. If yours has none, Zapier reaches thousands of other apps. - Connected in the conversation, guided step by step - Reads and, where allowed, writes in your tools - Every action visible in the chat - **Greenhouse:** Read candidates, roles and scorecards - **Workable:** Read candidates and move them between stages - **Ashby:** Read candidates and interviews, add notes - **Teamtailor:** Read and update candidates and jobs - **Calendar:** Book interviews and find slots, through a community connector - **Email:** Draft invites, feedback and offers, through a community connector - **Notion:** Keep role briefs and scorecards your team shares - **Slack:** Read hiring channels and post updates (with a token) (Shown: Interviews booked and invites drafted, from one sentence.) Example. A hiring manager asks how many candidates you’ve screened this month. You ask: “How many candidates did I screen in September, per role, and how many moved on?” Glowstone: Reads every interview note from September; Groups them by role; Counts who moved to the next stage. Result: A funnel per role in seconds, from your own notes. Also connectable (official or community connectors): Gem, HubSpot, Google Drive, Asana, Zapier, Linear. Dictation: speak instead of typing, to your agent or in any app, about three times faster. Free, with Whisper running on the Mac. ### A day with Glowstone - 09:00: Morning pipeline: who’s waiting on feedback, today’s interviews. - 10:00: Intake call recorded. Must-haves checked against your pipeline live. - 11:00: Interview recorded and filed into the candidate’s note. - 13:00: A profile captured from the tab, saved as a new candidate. - 16:00: Finalists compared side by side for the client. - Fri 17:00: A status per client, ready to send. ### Things to ask - “Who are my best three for this role, and why?” - “What did Tom say about his notice period?” - “Which candidates are waiting on feedback?” - “Compare the finalists on what the client cares about.” - “Draft the offer email for Irene.” - “Who in my pipeline has led a team?” ### Questions - **Do candidates need to know they’re recorded?** Yes. Recording people is regulated in many places; tell candidates and clients first and record only with their agreement. - **Where does candidate data go?** Your notes are files on your Mac; Glowstone has no server. The agent you choose reads what it needs to answer, under your account. Delete a candidate’s note and it’s gone. - **Does it replace my ATS?** No. The ATS keeps the process; Glowstone keeps the context: what each candidate said, what each client wants, how they compare. - **Is it free?** Yes. Glowstone, transcription and dictation are free. The agent uses your own plan with its provider. ## The AI second brain for customer success https://glowstone.ai/for/customer-success Every account, looked after. Glowstone is like an assistant who has read every call, ticket and note about your accounts: who uses what, what they asked for, what you promised at the last review. It thinks with the judgment of your best account manager and answers from your own records, so no renewal catches you by surprise. ### The problem - **Risk you see too late.** Usage drops, the champion leaves, a ticket goes unanswered. You find out at the renewal. - **QBRs built from scratch.** Every review means digging through calls, tickets and dashboards for an account you talked to three months ago. - **Promises across the team.** Sales promised a feature, support promised a fix, you promised a follow-up. Nobody has the full list. ### Keep every account, renewal and promise on record and ask your AI anything about them Each account, contact and renewal gets a note your agent writes and links as you work: health, usage, open issues, what each person cares about and every promise made to them. Then ask: which renewals are at risk and why, what you promised at the last QBR, who the real decision maker is. - Health, renewals and promises in one place - Every call and review linked to its account - With connectors, usage and billing join the answer (Shown: The renewals inside Glowstone, then the at-risk accounts and why.) Example. A QBR with Northwind tomorrow, and you last spoke three months ago. You ask: “Prepare the Northwind QBR: what changed, what we promised, what to raise.” Glowstone: Reads the account, the July QBR and open issues; Lists the promises and their status; Drafts an agenda with the risks first. Result: A one-page QBR prep: usage trend, the SSO promise, and the two questions to ask. ### Record customer calls and reviews and ask questions live, with the call and the account history as context Record check-ins, QBRs and escalations from the tab. No bot joins; Glowstone transcribes on your Mac. Mid-call, ask what the customer means by “last time”: what was promised, what their usage says, what support told them. Afterwards, Process updates the account and adds the follow-ups. - Promises and history, recalled mid-call - Usage and billing through connectors - Accounts updated after every call (Shown: A quarterly review: the exact promise and why the customer is upset.) Example. An escalation call with an angry customer. You ask: “Has Globex had this problem before, and how did we fix it?” Glowstone: Reads Globex’s past calls and issues; Finds the same error in March; Reads how it was resolved. Result: You tell them it happened in March and what fixed it, instead of starting from zero. ### Use it all day: capture dashboards, jot notes and ask from any app Most of the value comes from small moments, dozens of times a day. Glowstone hangs from the top of your screen in every app: bring the pointer up, ask, jot something down or attach what’s on screen, and get back to work. Away from the desk, a voice note to your own Glowstone bot on Telegram does the same. Everything lands in the same memory. - The tab, over any app - Screenshots it understands: “add this to…” - Telegram from your phone (Shown: A screenshot from the tab, filed into the right note.) - The tab. A customer pings you on Slack: “What’s the status of Northwind’s SSO request?” → Moved to November by product; you can answer in one line. - Attach a file. A customer sends feedback as a spreadsheet: “Group this feedback by theme and add it to Acme.” → Four themes with counts, saved to the account. - Telegram. After an on-site visit: “Voice note: “Hannah is leaving in December, Marc takes over.”” → Filed in Acme, and Marc flagged as the new champion. - Search ⌘F. Which accounts asked for an API?: ““API”” → Three accounts and the calls where they asked. ### Schedule routines: your AI sends you every Monday the accounts at risk and the renewals coming up Ask once: “every Monday at nine, tell me renewals in the next 60 days, accounts whose health changed and promises coming due”. It starts that conversation on its own. Silent when everything is green. Routines run while Glowstone is open on your Mac. - Renewals flagged two months ahead - Health changes, not just dates - Silent when all is well (Shown: Monday at 9:00, account health arrives on its own.) Example. You manage forty accounts and only hear from the loud ones. You ask: “Once: “every month, tell me accounts I haven’t spoken to in 60 days.”” Glowstone: Reads every account and its last call; Finds the quiet ones; Opens a list with a suggested check-in for each. Result: Nobody goes quiet without you noticing. ### Connect HubSpot, Stripe, Mixpanel and more, so your AI can read usage, billing and your CRM for you Most tools you work in today can be connected to an AI agent, through what are called MCP connectors. Tell Glowstone which ones you use and your agent sets them up with you, in the conversation: it looks for the official connector, adds it, and walks you through signing in. No settings screen, no developer needed. From then on it can read them and, where the tool allows it, act in them: create and update tickets, write pages, draft emails, check a customer’s invoices. Every action shows up in the chat, and it only reaches what your own account can. - Connected in the conversation, guided step by step - Reads and, where allowed, writes in your tools - Every action visible in the chat - **HubSpot:** Read accounts, contacts, deals and tickets (read-only connector) - **Salesforce:** Read and update records (official connector in pilot) - **Stripe:** Billing, plans and overdue invoices - **Mixpanel:** Usage and engagement by account - **Email:** Draft check-ins and follow-ups, through a community connector - **Calendar:** Your calls and reviews, through a community connector - **Slack:** Shared customer channels (with a token) - **Notion:** Account plans and playbooks (Shown: Usage from Mixpanel turned into three personal check-in drafts.) Example. Billing asks if an account is about to churn before they chase an invoice. You ask: “Is Globex behind on payments, and how are they doing as a customer?” Glowstone: Reads invoices through the Stripe connector; Reads the account’s health and last calls; Answers both in one line each. Result: One invoice late, but healthy and expanding. Billing can chase it gently. Also connectable (official or community connectors): Intercom, Amplitude, PostHog, Jira, Linear, Zapier. Dictation: speak instead of typing, to your agent or in any app, about three times faster. Free, with Whisper running on the Mac. ### A day with Glowstone - 09:00: Monday account health: renewals and risks. - 10:00: QBR prepared from the account’s history. - 11:00: Quarterly review recorded. A past promise checked mid-call. - 11:45: Process: the account updated, follow-ups listed. - 15:00: A usage dashboard captured and added to the account. - 17:00: Check-ins drafted for the accounts that dropped. ### Things to ask - “Which renewals are at risk, and why?” - “Prepare me for the Northwind QBR.” - “What did we promise Acme, and is it done?” - “Which accounts haven’t I spoken to in 60 days?” - “Who is the champion at Globex now?” - “Summarise the last three calls with Northwind.” ### Questions - **Does it replace our CS platform?** No. Your CRM or CS platform keeps the data; Glowstone keeps the context of every conversation and lets you ask across both through connectors. - **Does the customer see a bot in calls?** No. Glowstone listens on your Mac and works with any call app. Tell people you’re recording. - **Where does customer data go?** Your notes are files on your Mac; Glowstone has no server. The agent you choose reads what it needs, under your account. - **Is it free?** Yes. Glowstone, transcription and dictation are free. The agent uses your own plan with its provider. ## The AI second brain for freelancers https://glowstone.ai/for/freelancers Every client, brief and invoice, kept. Glowstone is like an assistant who has read every brief, call and email with your clients: what was agreed, what changed, what you quoted and what’s still unpaid. It thinks with the judgment of a seasoned freelancer and answers from your own records, so scope creep and late invoices stop costing you money. ### The problem - **Scope creep you can’t prove.** “Just one more page.” Was it in the quote? The brief changed three times and the details are in old emails. - **Invoices that slip.** You’re busy delivering, and an invoice sits unpaid for six weeks before you notice. - **Being your own assistant.** Calls, briefs, deadlines, admin. Every hour spent organising is an hour you can’t bill. ### Keep every client, brief and quote on record and ask your AI anything about them Each client, project, brief and quote gets a note your agent writes and links as you work: what was agreed, every change, the deadlines and what you’ve invoiced. Then ask: what changed since the original brief, whether a request is in the quote, what’s still unpaid. - Briefs, changes and quotes, linked - Deadlines and invoices in one place - Answers that cite the brief or email they came from (Shown: A project inside Glowstone, then the scope changes measured against the quote.) Example. A client says the new request was “always part of it”. You ask: “Show me where booking was mentioned in Café Lupa’s brief or calls.” Glowstone: Reads the brief, the quote and every call; Searches for booking or reservations; Quotes what was actually said. Result: It was never mentioned; the quote covers five pages. You reply with facts, politely. ### Record client calls and ask questions live, with the call and the client’s history as context Record kick-offs and feedback calls from the tab. No bot joins, so the client sees nothing different; Glowstone transcribes on your Mac. When a client asks for something new, ask the tab whether it’s in the quote before you say yes. Afterwards, Process files the feedback and any change of scope into the project. - Scope checked before you say yes - Feedback filed into the project - Every change of scope on record (Shown: A feedback call: “just one more page” checked against the quote, live.) Example. After a long kick-off call with a new client. You ask: “Turn today’s kick-off into a brief and a list of open questions.” Glowstone: Reads the call transcript; Writes the brief in the client’s words; Lists what wasn’t decided. Result: A brief to send for approval the same afternoon, with the five questions that would have bitten you later. ### Use it all day: capture designs, invoices and pages, and ask from any app Most of the value comes from small moments, dozens of times a day. Glowstone hangs from the top of your screen in every app: bring the pointer up, ask, jot something down or attach what’s on screen, and get back to work. Away from the desk, a voice note to your own Glowstone bot on Telegram does the same. Everything lands in the same memory. - The tab, over any app - Screenshots it understands: “add this to…” - Telegram from your phone (Shown: A screenshot from the tab, filed into the right note.) - The tab. A client emails a vague request: “Is this in the Studio Nord quote?” → No: templates were capped at ten. You reply with a change order. - Attach a file. A client sends feedback as a PDF with comments: “Turn these comments into a task list for the homepage.” → Twelve tasks, grouped by section, in the project note. - Telegram. An idea while walking: “Voice note: “Café Lupa logo could use the tile pattern from their floor.”” → Filed in the Café Lupa project. - Search ⌘F. What did you charge for a similar job?: ““brand book”” → Two past quotes and what you charged each time. ### Schedule routines: your AI sends you every Friday unpaid invoices and next week’s deadlines Ask once: “every Friday at five, tell me unpaid invoices, deadlines next week and hours I haven’t billed”. It starts that conversation on its own. Silent when everything is paid and on time. Routines run while Glowstone is open on your Mac. - Unpaid invoices never forgotten - Deadlines across every client - Silent when there’s nothing to chase (Shown: Friday at 17:00, money and deadlines arrive on their own.) Example. January, and you need your year in numbers for taxes. You ask: “Once: “on the 2nd of each month, summarise what I invoiced and what was paid.”” Glowstone: Reads your invoices note every month; Adds up invoiced, paid and overdue; Opens a short summary. Result: Monthly numbers ready for your accountant, without a spreadsheet session. ### Connect Stripe, Xero, Notion and more, so your AI can invoice, track and organise for you Most tools you work in today can be connected to an AI agent, through what are called MCP connectors. Tell Glowstone which ones you use and your agent sets them up with you, in the conversation: it looks for the official connector, adds it, and walks you through signing in. No settings screen, no developer needed. From then on it can read them and, where the tool allows it, act in them: create and update tickets, write pages, draft emails, check a customer’s invoices. Every action shows up in the chat, and it only reaches what your own account can. - Connected in the conversation, guided step by step - Reads and, where allowed, writes in your tools - Every action visible in the chat - **Stripe:** Payments, invoices and who has paid - **Xero:** Accounting: invoices, payments and expenses - **Figma:** Read designs and specs for client work - **Notion:** Client portals, briefs and project pages - **Google Drive:** Find and read client files and deliverables - **Email:** Draft replies and reminders, through a community connector - **Calendar:** Calls and deadlines, through a community connector - **GitHub:** Repos and issues for development work (with a token) (Shown: Hours from your notes turned into a draft invoice in Xero.) Example. A client asks if their last payment arrived. You ask: “Has Studio Nord paid invoice #041?” Glowstone: Checks the payment through the Stripe or Xero connector; Reads your invoices note; Answers with the date. Result: Paid on Sep 3. You answer in ten seconds, without opening your bank. Also connectable (official or community connectors): Asana, monday.com, Canva, Linear, Zapier. Dictation: speak instead of typing, to your agent or in any app, about three times faster. Free, with Whisper running on the Mac. ### A day with Glowstone - 09:00: Today’s deadlines across every client. - 10:00: Kick-off call recorded, turned into a brief. - 12:00: A client’s request checked against the quote from the tab. - 16:30: Feedback call. “Just one more page” checked live. - 17:00: Process: feedback and a change of scope filed. - Fri 17:00: Unpaid invoices and next week’s deadlines. ### Things to ask - “Is this request in the quote?” - “What has this client changed since the brief?” - “Which invoices are unpaid, and for how long?” - “What did I charge for a similar project?” - “Turn this call into a brief.” - “How many hours have I worked for Acme this month?” ### Questions - **Can I keep each client separate?** Yes. Each client has its own notes, or use separate spaces: each is its own folder with its own conversations. - **Does it send invoices for me?** With a connector like Xero or Stripe it can draft them. You review and send; it doesn’t send anything you haven’t approved. - **Where do my clients’ files go?** Your notes are files on your Mac; Glowstone has no server. The agent you choose reads what it needs, under your account. - **Is it free?** Yes. Glowstone, transcription and dictation are free. The agent uses your own plan with its provider, or free models with OpenCode or Kilo Code. ## The AI second brain for customer support https://glowstone.ai/for/customer-support Every answer, every precedent. Glowstone is like a senior colleague who has read every help article, every customer record and every tricky case your team has ever resolved. It thinks with the judgment of your most experienced support lead and answers from your own records, so you can decide edge cases yourself and reply well, in the customer’s own language. ### The problem - **Edge cases wait for someone else.** A refund outside policy, an exception for a big account. You escalate to a manager or product, and the customer waits a day. - **The same question, answered ten ways.** How the team handled a case last month lives in someone’s head or a Slack thread nobody can find. - **Context scattered across tools.** The ticket in the help desk, the payment in Stripe, the policy in a doc, the customer’s history in the CRM. ### Keep your help docs, customer records and past decisions on record and ask your AI anything about them Your policies and help articles, a note per important customer, and a record of how ambiguous cases were resolved and why: a body of precedent your agent writes and links as the team works. Then ask it before you escalate: has this happened before, what did we decide, what does the policy actually say. You stop depending on a manager for every exception, because the reasoning is already on record. Working as a team? Keep the vault in a shared Google Drive or Dropbox folder and point everyone’s Glowstone at it: the policies, precedents and customer notes are the same for the whole team, and a ruling one person saves is there for the next. And if you run support on your own, as a founder or a product manager, it’s your whole support memory in one place. - Policies and help articles in one place - Shareable with your team through a Drive folder - Past rulings on edge cases, with the reason - Answers that cite the policy or precedent (Shown: A precedent inside Glowstone, then an edge case decided without escalating.) Example. A new kind of request no one on the team has seen. You ask: “A customer wants to pause their annual plan for two months. Have we ever done that?” Glowstone: Searches policies and every precedent; Finds two similar cases and how they ended; Says what applies and what’s still a manager’s call. Result: We offered a free month instead of a pause, twice. You can offer the same, and you know exactly when to escalate. ### Record customer calls and team handovers and ask questions live, with the call and the customer’s history as context Record a phone or video call with a customer, or the morning handover with your team, from the tab. No bot joins; Glowstone transcribes on your Mac. Mid-call, check the customer’s history, the policy or a past ruling without putting them on hold. Afterwards, Process files the outcome in the customer’s note, and a new decision becomes precedent for the next person. - Customer history and policy, checked mid-call - New decisions saved as precedent - Handovers written down, not lost (Shown: A customer call: the charge checked in Stripe and the precedent applied, live.) Example. The night shift hands over 40 open tickets. You ask: “From this handover, which tickets are urgent and what was promised to each customer?” Glowstone: Reads the handover transcript; Matches each ticket with the customer’s note; Lists the promises and deadlines. Result: Six urgent tickets and three promises due before noon, before you open the queue. ### Use it all day: draft replies in any language, capture tickets and ask from any app Most of the value comes from small moments, dozens of times a day. Glowstone hangs from the top of your screen in every app: bring the pointer up, ask, jot something down or attach what’s on screen, and get back to work. Away from the desk, a voice note to your own Glowstone bot on Telegram does the same. Everything lands in the same memory. Drafting replies is where it saves the most: you write what you want to say, in your words, and it turns it into a clear, kind answer in the customer’s own language, with the right policy behind it. - The tab, over any app - Screenshots it understands: “add this to…” - Telegram from your phone - Replies in the customer’s language (Shown: A screenshot from the tab, filed into the right note.) - The tab. A customer writes in Portuguese: “Reply in Portuguese: we can’t refund, but we can offer a free month.” → A warm, clear answer in Portuguese, with the right policy behind it. - Screenshot gesture. A customer sends a screenshot of an error: “↓ + → and “what is this, and is there a known fix?”” → The known issue, its workaround, and the help article to link. - Attach a file. A new policy arrives as a PDF: “Summarise what changes for refunds and save it.” → The changes saved into Refund policy, with the date. - Search ⌘F. Has anyone handled this before?: ““partial refund agency”” → The precedent and the ticket where it was decided. ### Schedule routines: your AI sends you every morning the tickets at risk and new precedents Ask once: “every morning at eight, tell me tickets about to breach their deadline, promises due today and any new precedent the team saved yesterday”. It starts that conversation on its own. Silent when there’s nothing urgent. Routines run while Glowstone is open on your Mac. - Deadlines flagged before they’re breached - The team’s new decisions, shared every day - Silent when the queue is calm (Shown: 8:00, the support brief arrives on its own.) Example. The same question keeps coming in. You ask: “Once: “every Friday, tell me the top five reasons customers contacted us this week.”” Glowstone: Reads the week’s handovers and notes; Groups contacts by reason; Opens a short report with examples. Result: A weekly report for product, with real examples, written for you. ### Connect Zendesk, Freshdesk, Stripe and more, so your AI can read tickets, payments and customer records for you Most tools you work in today can be connected to an AI agent, through what are called MCP connectors. Tell Glowstone which ones you use and your agent sets them up with you, in the conversation: it looks for the official connector, adds it, and walks you through signing in. No settings screen, no developer needed. From then on it can read them and, where the tool allows it, act in them: create and update tickets, write pages, draft emails, check a customer’s invoices. Every action shows up in the chat, and it only reaches what your own account can. - Connected in the conversation, guided step by step - Reads and, where allowed, writes in your tools - Every action visible in the chat - **Zendesk:** Read tickets and customers, through community connectors - **Freshdesk:** Tickets and contacts, through Freshworks’ MCP gateway - **Help Scout:** Conversations and customers (official connector) - **Intercom:** Conversations and user data - **Stripe:** Payments, refunds, subscriptions and invoices - **Shopify:** Orders, shipping and returns - **HubSpot:** Contacts, companies and tickets (read-only connector) - **Notion:** Your help centre and internal playbooks (Shown: Help desk tickets checked against Stripe, one by one.) Example. A customer asks where their order is and you don’t have access to logistics. You ask: “Where is order #4471, and has it been delayed before?” Glowstone: Reads the order through the Shopify connector; Reads the customer’s note; Checks past delays for the same carrier. Result: In transit, one day late; the carrier has had delays all week. You answer with the tracking and an honest estimate. Also connectable (official or community connectors): Salesforce, Slack, Jira, Linear, Google Drive, Zapier. Dictation: speak instead of typing, to your agent or in any app, about three times faster. Free, with Whisper running on the Mac. ### A day with Glowstone - 08:00: The support brief: tickets at risk and yesterday’s new precedents. - 09:00: An edge case decided from a precedent, no escalation. - 10:40: Customer call recorded; the double charge checked in Stripe live. - 11:00: Process: the outcome saved, and a new precedent for the team. - 14:00: Replies drafted in German, Italian and Portuguese. - Fri 17:00: Top contact reasons of the week, for product. ### Things to ask - “Can I approve this refund, and what does the precedent say?” - “Has a customer asked for this before? How did we handle it?” - “Reply in French: we can offer a free month instead.” - “What did we promise this customer last time?” - “Which tickets are about to breach their deadline?” - “Summarise this customer’s history before I call.” ### Questions - **Does it replace our help desk?** No. Tickets stay in Zendesk, Freshdesk or wherever you use. Glowstone keeps the knowledge around them: policies, precedents, customer notes, and lets you ask across all of it, including the help desk through connectors. - **Can the whole team share the precedents?** Yes. The vault is a normal folder, so put it inside a shared Google Drive or Dropbox folder and point each person’s Glowstone at it (Settings → Change folder). Policies, precedents and customer notes are then shared by everyone. Conversations stay on each person’s Mac. There’s no locking: if two people edit the same note at the same moment, the sync service keeps both versions, so it works best when each precedent is written once and then read. - **Does it send replies to customers?** It drafts them. You read, adjust and send from your help desk. Nothing goes to a customer without you. - **Where does customer data go?** Your notes are files on your Mac; Glowstone has no server. The agent you choose reads what it needs, under your account. Check your company’s data policy before adding customer records. - **I’m a founder or PM handling support myself. Is it for me?** Very much. When one person runs support, the hard part is remembering every customer and every exception you’ve made. Glowstone keeps both, drafts replies in each customer’s language, and with Stripe or your help desk connected answers most questions without opening another tab. See also the pages for founders and product managers.