Meeting notes for Windows
Record the call.
No bot. No cloud.
KeptInFlow records Teams, Zoom, Meet, and Slack huddles from your own PC, tells the speakers apart, and writes the summary. Nothing joins your meeting, and with the free on-device models no audio leaves your machine. Pay once, keep forever.
If Windows SmartScreen asks about this new release, click More info → Run anyway. The installer is code-signed and verified.
Your mic and the call audio, captured locally. The meeting never sees it.
One install replaces
- Meeting notetaker
- Work journal
- Dictation app
- Snipping Tool
- Screen recorder
- Clipboard manager
- ChatGPT tab
- Text expander
Runs on your PC, not in a tab · no account · pay once
Every call, searchable
Record the meeting.
Ask it questions later.
One shortcut (or auto-record) captures your mic and system audio, transcribes with each speaker told apart, and writes the briefing. Weeks later, search every meeting you've ever recorded and play the exact second a thing was said.
Why not just use a notetaker bot?
Every other tool solves this by sending a robot into your meeting. That one design choice is where all the friction comes from.
| Bots that dial in | KeptInFlow | |
|---|---|---|
| In the participant list | One extra attendee, visible to everyone in the room | Nobody new. There is nothing to explain |
| To start recording | Hand over your calendar, or paste the link and wait for the host to admit it | One shortcut, or it notices the call and offers |
| Where your audio goes | Up to their servers, because that is where it gets transcribed | Can stay on this PC from recording to transcript |
| A call nobody scheduled | Only if you catch it in time and paste the link | Already recording |
| Audio that is not a meeting | Not what a meeting bot is for | A webinar, a phone on speaker, an interview: same shortcut |
| If your company bans bots | You are out of luck | There is no bot to ban |
| What it costs | Per seat, every month, for as long as you use it | Once |
The trade is real: a bot can sit in a meeting you never joined. KeptInFlow records what your own PC can hear, so you have to be in the call.
Auto-record, or one key
KeptInFlow can notice when Zoom, Teams, a Slack huddle, Webex, Discord, or a browser call starts using your mic, then ask or just start recording. Off by default, stops itself when the call ends, and you pick which apps count.
Who said what
Each voice gets a name and a colour. Rename someone once and the next recording suggests them, one click to accept. A high-accuracy model handles calls that noise removal has flattened, and a misheard name is one click to fix across the transcript.
Summary, action items, answers
One click writes the overview, what was decided, and who took what, every line carrying the timestamp it came from. The Ask tab answers "what did I miss?" from the transcript alone, citing the moments it read.
Search every meeting at once
Type a word and every transcript, title, and summary is read together. Pick a line and that meeting opens with the hit highlighted and the audio cued to the second it was said.
Transcription and speaker identification can run entirely on this PC with free downloaded models, so a sensitive call never has to leave it. Recordings shrink to about a tenth of their size once transcribed, and playback can skip the silence.
What Windows users ask for most
Copy text from anything.
Find any screenshot later.
On-device OCR (the same class of engine Windows uses) plus a capture library you can search by the words that were on screen, not by filename. No cloud upload for OCR.
Copy text off your screen
Snip a region. Text lands on the clipboard. AI cleanup when the layout is messy.
Search past screenshots by content
Every snip is indexed. Type words that were on screen. Images become findable.
Snipping, OCR, annotation and the searchable Library have a page of their own.
Works in every app you already use
Stop alt-tabbing to your AI.
System-wide tools for Outlook, Word, Slack, your IDE, the browser: anywhere you can type. Each one is a single trigger away.
The routine you know
- Alt+Tab to the browser
- Find the AI tab
- Paste, then retype the context
- Wait. Copy the answer
- Alt+Tab back, find your place
- Paste, fix the formatting
With KeptInFlow
- Type the trigger where you are
The answer lands where your cursor already is. Your train of thought never leaves the station.
Works for me. Flights land at 11:40am New York time, so let's say noon.
Typed in any app. The answer lands in place.
Answers in place
Type a question between << >> mid-sentence and the
answer replaces it where you're typing. Add ? to search the
web, or say screen to have it read the window you're in.
Edit by meaning
Select text, say what to change (“make this friendlier,” “turn into bullets”), and review the diff before it lands. Streaming, undoable, with quick-action presets.
Dictate anywhere, offline
Hold, speak, release. Cleaned-up text pastes at your cursor. With the free on-device speech model, dictation is unlimited, works with no connection, and your voice never leaves this PC. Use Alt+Shift+A to speak an instruction instead and it runs as an AI request. See dictation.
Clipboard cycler
Alt+Tab for your clipboard: hold the modifier, flip through recent clips, release to paste the one you meant. Pin clips, jump by number, cycle text case. See the clipboard manager.
Also system-wide
Verification codes, caught
Click into a one-time-code box and the code is already sitting under it, with Tab to take it. It finds the code even when it landed in a window buried behind your browser, or one you never switched to.
Fetch
Click into a box asking for an order number, an email or a phone number, and the thing you copied that fits is offered right above it. Tab takes it. Alt+Shift+W asks in any box at all, even one with no label.
Smart paste
Fills the whole form in front of you from details you already have, matching each box by what it is asking for. You see a preview before a single character is written.
Text expansions
Abbreviations that expand in any app, with templates, fill-in
forms, and AI expansions where the text is a prompt and the reply
is what pastes. The ;; picker searches them, creates
new ones in place, and can just answer a typed question.
See text expansion.
AutoCorrect
System-wide typo fixing as you type, in every app. One Backspace undoes any correction. Alt+Shift+F rewrites your last sentence properly.
Calculator door
Type math between = = and the result replaces it.
Units, dates, and word problems included. No app switch for
arithmetic ever again.
Clean paste
Paste as tidy plain text: formatting stripped, whitespace fixed. A copied spreadsheet range pastes as a Markdown table. Instant, no AI call, no key needed.
Codes, Fetch, and whole-form filling have a page of their own.
Don't take our word for it
Type one yourself.
This box simulates the mechanic. Type math between
= = or a question between << >>
and watch it resolve in place. In the app it works exactly like this,
in every text box on your PC, answered by your own model.
A simulation with canned answers. The real app sends your question to the AI model you choose.
Nothing to start. Nothing to tag.
Friday, five o'clock.
Your week is already written.
Turn on Screen Memory and every day writes itself up: what you finished, what is still going, what got stuck. Scrub back to any moment, read a week back before a standup, or ask one question across everything you have seen, said, and saved. All of it on this PC.
The day writes itself down
No timer to start and nothing to tag. Open a day and it is already written: what finished, what is still going, what is stuck, and the stretches of work it broke into. Weeks and months get the same write-up, one button from a standup. See the work journal.
Scrub back to any moment
A timeline of your screen you can search by the words that were on it, or by meaning: "error about permissions" finds the stack trace. Filter by app, click a frame, copy the text off it.
One question, every source
"What did I decide about pricing?" is answered from everything you have seen, said, and saved, with each moment cited so you can check it.
Off by default, and honest about itself: capture and recall stay on this PC, only the matched text ever goes to your AI model, and the main window always shows what is running in the background, one click from off.
One box for the whole PC
Type what you want. It happens.
A command palette for all of Windows: launch any app, switch to any open window, find your notes, snips, meetings, and files, run a command, or just ask a question, all from one shortcut.
Launch or switch, instantly
Every installed app and every open window, a few letters away. A running app's row switches to it; Ctrl+Enter starts a second copy. It learns which result you mean and leads with it next time.
Search everything you've kept
Scope a search to your snips, notes, clipboard, meetings, or files, and ask full questions of them: "what did we decide about pricing?" answers from the matches, sources cited.
Run, kill, convert, control
Run any command like Win+R, end a frozen app by name, set the volume, convert units and currencies, or give a pin a keyword so "yt cats" searches YouTube. No mouse in any of it.
Ask anything from anywhere
Type a question and the answer streams into a floating note, or is written straight into the app you came from. On models that can, it checks the web when the question needs a live answer.
Serious screen capture, built in
A better snipping tool that feeds AI.
Precise capture, annotate, pin, and record. Everything you snip is one step from an answer, and stays searchable in the Library.
Snip with real OCR
Drag a region with precise handles, a loupe, and live dimensions. Copy as image or text. On-device OCR with optional AI cleanup, better than PowerToys for messy layouts. Or snip with just the mouse: hold right, click left, drag.
Searchable Library
Every capture is archived and OCR-indexed. Find a screenshot by the words that were on it, days later, without digging through folders.
Annotate & redact
Arrows, boxes, numbered steps, blur, spotlight, crop, plus one-click auto-redaction that finds emails, card numbers, and keys on-device and blurs them.
Record, then edit in place
Record a region to GIF or H.264 MP4, then edit it right there: speed up the dull parts, cut dead air, blur secrets, zoom into what matters. Trim with a live preview. No separate editor needed. See screen recording.
Scrolling capture
Capture a long page end-to-end: KeptInFlow scrolls the window and stitches the frames into one tall, clean image.
The part nothing else does
Then there's the Deck.
Launcher assistants answer one thing and vanish. KeptInFlow lets you gather material from several windows, run one prompt across the whole pile, and keep the answer floating on screen while you work.
Collect
Snip text and screenshots from different windows (a doc, an email thread, a code file) into one shared pile. A hotkey or mouse gesture away.
Synthesize
Run one prompt across everything you gathered: “find the contradiction,” “turn this into an explainer,” “draft the reply.”
Pin
The answer becomes a floating, always-on-top note. Drag it, resize it, keep the conversation going. As many notes as you want.
Notes can watch a file, URL, or window and refresh themselves; link notes so one transforms another's answer; lay the whole thing out on a spatial canvas.
Built for the skeptical
Private by architecture.
Your keys, your data
Bring your own OpenAI, Anthropic, Google, xAI, or OpenRouter key. Keys live in the Windows keychain, are read only by native code, and never touch a server we run. With your own key, prompts go straight from your machine to your provider. The free monthly actions are the one exception: those pass through our relay, and we mask detected personal details before forwarding them.
No account, no backend
Nothing to sign up for. Conversations, captures, and settings are local files on your disk. Optional on-device masking swaps emails, card numbers, and keys for placeholders before a prompt is sent, puts the real values back in the answer, and shows you exactly what was hidden. Screenshots get their own version: one click blurs detected personal details before you share.
The heavy lifting is local
OCR, capture search, Screen Memory, dictation, transcription, and speaker identification can all run on this PC with free downloaded models. Background capture is opt-in, and the main window always shows what is running, one click from off.
Honest cost tracking
A real usage dashboard: per-provider spend, token counts, and web search costs, measured from actual responses, not estimated.
Most apps this capable ask you to take that on faith. Instead, here is the whole list: every feature that reads anything, what it reads, whether it leaves your PC, and whether it is on when you install. See what it watches
From download to first answer
Set up in two minutes.
Install
A signed Windows installer. No sign-up, nothing to configure. Setup asks for your email so the free actions are yours.
Hit a shortcut
It works immediately. 1,000 free AI actions come with the email you entered, so you can try everything without an API key.
Add your key when ready
Paste a key from OpenAI, Anthropic, Google, xAI, or OpenRouter. KeptInFlow detects the provider and stores it in the Windows keychain.
No subscription required
Pay once. Own it.
Try every feature free for 14 days. No card to start.
App forever. Updates for 1 year. Keep the last version you got if you don't renew. This is not a stealth subscription. Details
What it replaces, at their prices
| Included in KeptInFlow | Sold separately as | Their price |
|---|---|---|
| Meeting recording, transcription, speaker labels | Krisp, Otter | $96 to $100/yr |
| System-wide dictation, on-device, unlimited | Wispr Flow Pro | $144/yr |
| Text expansion everywhere | TextExpander | $40/yr |
| Screen capture, OCR, annotation | Snagit | $39/yr, no perpetual option |
| Screen recording, polished and shared | Loom Business | $96/yr a person |
| A work journal that writes itself: days, weeks, what you finished | Timely, Rize, RescueTime | $108 to $144/yr |
| Screen Memory: search everything you've seen | Rewind and Limitless shut down. Windows Recall needs a Copilot+ PC. | |
Over $500 a year of subscriptions, replaced with one purchase. Prices are annual billing, checked July and August 2026. Check them yourself; they move.
Free trial
14 days
- Every feature, no limits
- Bring your own API key
- 1,000 free AI actions included to try it without a key
- No card up front. Nothing to cancel.
License
$99 $59 one-time
Founding price for the first 100 · 1 year of updates included
- Own it forever, no subscription
- 1 year of updates included
- Renew updates later for $29/yr, or don't, and keep what you have
- Unlimited on-device dictation, no quota to run out of
- 30-day refund, no questions asked
- $229 lifetime-updates option
$59 for the first 100 buyers. Then $79 for the next 200. Then $99. Each step ends when it sells out, not on a timer.
Want every future update instead of one year? Lifetime updates, $229. Both are a single payment, and both come with the 30-day refund.
Don't want to deal with an API key? There is an optional plan that handles the AI for you: $12/month, no key to manage, 5,000 fast actions and 1,200 minutes of cloud dictation a month. It unlocks the full app while it runs, so it works on its own or on top of a license. On-device dictation and transcription stay unlimited either way.
Before you ask
Questions, answered.
Does a bot join my call?
No. KeptInFlow records from your PC: your microphone and the sound your speakers are playing. Nobody else in the meeting sees an extra participant, and there is nothing for them to consent to or object to. This is the main reason people pick it over the notetakers that dial in, and it is usually the difference between "allowed" and "blocked" at companies that ban meeting bots.
Does my audio go to the cloud?
It does not have to. Transcription and speaker identification can run entirely on your PC using free models you download once, so a sensitive call never leaves the machine. Writing a summary or asking a question about a meeting is the step that calls an AI provider, and that only happens when you click it. You can record and transcribe all day without any of it leaving your PC.
What does KeptInFlow watch when I'm not using it?
Less than you would guess, and we list every bit of it. The background screen history is off unless you turn it on, auto-recording is off unless you turn it on, and the small on-screen offers read the labels of the field you are typing in, not the contents of your screen. The full breakdown, feature by feature, is on the what it watches page: what each one reads, whether anything leaves your PC, and how to switch it off.
Do I need an API key?
Not to try it. The trial includes 1,000 free AI actions with no key of your own, unlocked by the email you enter in setup, and they can search the web for live answers. For everyday use you add your own key from OpenAI, Anthropic, Google, xAI, or OpenRouter; paste it once and KeptInFlow detects the provider. You pay your provider directly at their rates, typically pennies a day for these workloads.
Is this a subscription?
No. KeptInFlow is a one-time license: the app is yours forever, with a year of updates included. Renewing updates later is optional ($29/yr), versus the $39 every year that a typical capture-tool subscription costs. If you don't renew, you keep the last version you received. You are not locked out. The optional $12/month AI Included plan is for people who don't want to manage an API key. It also unlocks the app while it runs, so you can subscribe instead of buying, but you never need it to use the app you bought.
Why one-time, when everything else is monthly?
Because your dictation and screen history run on your machine with your keys. We don't carry per-user cloud costs for the core app, so we don't need rent. The optional $12/month plan exists to cover AI costs for people who'd rather not manage an API key. And if it doesn't stick in your workflow, there's a 30-day refund, no questions asked.
Does OCR need the cloud?
No. Text extraction runs on your machine (Windows OCR / OneOCR). An optional AI cleanup pass only runs if you choose it and only sends the extracted text to the provider you configured. Searching your capture Library is local.
Where does my data go?
With your own key: from your machine directly to the AI provider you chose, nowhere else. No KeptInFlow server in the path, no account, no telemetry-driven backend. Keys are stored in the Windows keychain and read only by native code. Optional on-device masking swaps emails, card numbers, and keys for placeholders before sending, and restores them in the answer. Every feature that reads anything is listed on the what it watches page.
The 1,000 included actions work without a key, and those do pass through a KeptInFlow relay so we can talk to the model on your behalf. Detected personal details are always masked on that path, whatever your setting says. Add your own key and the relay leaves the picture.
Which models can I use?
GPT, Claude, Gemini, and Grok families, plus anything on OpenRouter. Switch per note or per feature, with web search on models that support it. Your default is one click to change.
Does it really work in every app?
The typing tools work anywhere Windows lets you type: email clients, browsers, IDEs, chat apps, terminals. KeptInFlow types the result in like you would, so there's no per-app integration to break.
What about macOS?
Windows first: that's where the deepest integration is possible today. macOS is on the roadmap but not dated; the license will cover both when it arrives.
Record the call. Keep the notes. Own the app.
If Windows SmartScreen asks about this new release, click More info → Run anyway. The installer is code-signed and verified.