NewAnnotate Live is here

Don't type.
Just point and speak.

Stop describing your screen in paragraphs. Record it, circle what you mean and say what you want, and your AI gets to work.

  • Free forever
  • No sign-up
  • Stays on your computer
  • Mac & Windows
Settings · your-appDocs
your-app.com/settings

Profile settings

Full name
Email
Bio
CancelSave changes
Claude3 fixes ready to review
Settings redesign · Mockups
Layers
Mobile
Profile card
your-app — npm run dev
~/your-app npm run dev
▲ Next.js · Local: http://localhost:3000
✓ Ready in 1.2s
GET /settings 200 in 84ms
POST /settings 500 in 31ms
⨯ TypeError: Cannot read properties of undefined (reading 'email')
at saveProfile (app/settings/actions.ts:14:22)
at async POST (app/settings/route.ts:9:3)
~/your-app
BrowserMockupsTerminalClaudeFileEditView9:41

“Make this button green, like the sidebar.”

“Use this card layout from the mockup.”

“And fix this error when I hit Save.”

Recording saved3 apps · 0:24

Send to

ClaudeCursorCodex
RecordStop
Claude
Annotate recording · 3 apps · 0:24Make the changes I showed in this recording.
I watched your recording. You marked 3 things:
Browser: Save button, green like the sidebar
Mockups: use this profile card layout
Terminal: TypeError on Save
Edited profile-form.tsx+1 −1
- <Button variant="ghost" size="sm">
+ <Button className="bg-primary px-5">
Done. 3 fixes are ready for you to review.
Reply to Claude…
  1. 1Hit record
  2. 2Point and talk
  3. 3Send to your AI
  4. 4It's fixed

Works with the AI tools you already use

  • Cursor
  • Claude Code
  • Codex
  • Claude Desktop
  • + any app that supports MCP

Why Annotate

Explaining a screen in words is the slow part.

You end up typing a paragraph to describe one button, and the AI still guesses which one you meant.

Typing it out67words

On the settings page, the button at the bottom right, the one that says Save changes, not Cancel. It's grey and squished right up against Cancel. Can you make it the same green we use in the sidebar? I think it's the active nav colour, maybe a token? And give it more padding, but don't change the Cancel one. Also it should still look fine on mobile…

67 words, and the AI still has to guess.
Showing it with Annotate6words
CancelSave changes

“Make this green like the sidebar.”

  • The exact spot
  • What you said, word for word
  • Every click and screen change
Voice-only (Whisper Flow etc.)27words
Finally done

“Change the button — not Cancel, the other one at the bottom right, the grey one, next to Cancel… the Save one? Not sure what it's called.”

Can't see your screen — still need to describe exactly where
With Annotate3words
CancelSave changes

“Make this green.”

No name needed — just point at it

Who it's for

Your job, minus the explaining.

Pick your role and see what a short recording replaces.

For developers

Skip the prompt. Ship the fix.

The problem. You write a long prompt to explain which screen, which button and which error. The AI still edits the wrong file.

  • Circle the bug and say what's wrong. No prompt to write
  • The AI sees your app and your terminal in the same recording
  • Opens in Cursor, Claude Code or Codex, where you already work
Try it free

How it works

Three steps. Zero prompt writing.

If you can share your screen on a call, you already know how to use it.

  1. 1

    Hit record

    One click or a shortcut. Annotate records the screen your cursor is on, and follows it when you move to another.

    Screen 1
    Screen 2
    Recording · 0:07
  2. 2

    Point and talk

    Circle, box or underline anything on screen. Say what you want, the way you'd tell a teammate.

    “This card looks off. Match the spacing of the other two.”

  3. 3

    Your AI takes it from there

    Cursor, Claude Code or Codex opens the recording, sees exactly what you showed, and starts the work.

    ClaudeNew recording
    Key frames from your screens
    Your marks and clicks
    What you said, with timestamps
    Got it. Matching the middle card's padding to the other two and updating PlanCard.tsx.

See it for real

Watch a real session, start to finish.

One recording, handed to an AI, turned into working code.

New · Annotate Live

Talk it through. It sees what you see.

A voice partner that understands your screen while you record. Ask a question, point at something, and it acts on it with you.

  • Answers questions in real time about what's on your screen
  • Sends code changes to your coding agent
  • Tells you out loud when the work is done
Included in Starter and Pro
Annotate LiveSees your screen

“Why is this chart empty for last week?”

“The date filter on the left is set to this month only. Want me to have your coding agent default it to the last 30 days?”

“Yes, do that.”

Task sent: default the date filter to the last 30 days

Cloud transcription, in your language

A frontier speech-to-text model in 20 languages. Pick yours, save faster, and keep your computer free.

  • English
  • Spanish
  • French
  • German
  • Portuguese
  • Italian
  • Dutch
  • Japanese
  • Korean
  • Chinese
  • Hindi
  • Arabic
  • Russian
  • Turkish
  • Vietnamese
  • Indonesian
  • Filipino
  • Thai
  • Polish
  • Ukrainian

Cloud recordings for cloud agents

For agents that run in the cloud instead of on your computer. Upload a recording and they can read it too. Your local copy always stays.

  • Claude Code on the web
  • Cursor cloud agents
  • Lovable
  • v0
  • Base44
  • Figma
  • + any agent that supports MCP

Features

Everything your AI needs to get it right the first time.

Annotate captures the context you'd otherwise spend minutes typing.

Session handoff

Stop recording. Your agent is already on it.

Run /annotate-session-start once in your agent chat and it shows up on the Annotate toolbar. Click it after a recording, and the work starts right there. No copying, no pasting.

  • One click from recording to work started
  • Keep several agent chats listening at once
  • Each chat shows its latest change on the toolbar
  • Works with cloud agents too, with Cloud on

Mark the exact spot

Pen, box and text tools, so there's no doubt which button you mean.

0:03 “this one” · 0:08 “make it green”

Your voice, synced to the screen

Everything you say is transcribed and matched to what was on screen at that moment.

Every screen, one take

App on one display, code on the other. Both captured together.

Light on tokens

Your AI reads the key frames and your words, not a heavy video file. Faster and cheaper.

CursorClaude CodeCodex

Opens where you already work

Cursor, Claude Code or Codex. No new app to learn, no copy-pasting.

Cloud agents too

Turn on Cloud, and agents running in the cloud read your recordings through Annotate's remote MCP.

Private by design

Your screen stays on your machine.

The cloud is there if you want it, and only if you want it. It lets AI agents running in the cloud read a recording they can't reach on your computer.

Read the Privacy Policy
  • Local by default

    Recordings, frames and transcripts are saved on your computer.

  • Cloud is opt-in

    Nothing uploads unless you switch Cloud on yourself.

  • Cloud copies expire

    Uploads are deleted after 30 days. Your local copy is never touched.

Pricing

Free stays free.

Everything you need to record and hand off is free. Upgrade only for cloud extras.

Free

$0/ month

Unlimited local Whisper transcription in English, and Live on your own Gemini key. 1 hour of cloud recordings when signed in.

Starter

$5/ month

3,000 min of cloud transcription in your language, 10 hours of cloud recordings, and 30 min of Live, no key needed.

Pro

Most power

$15/ month

Unlimited cloud transcription and cloud recordings, 2 hours of Live, and prepaid Live credits when you want more.

FAQ

Questions, answered.

Do I need to know how to code?
No. You record, point and talk. Annotate turns that into something an AI coding assistant can read. Connecting it to your AI tool is a one-time copy and paste, and there's a step-by-step guide.
Which AI tools does it work with?
Cursor, Claude Code, Claude Desktop and Codex, plus any app that supports MCP, such as Lovable, Figma, Base44 and v0. Tools that run in the cloud read your recording through Cloud. See the setup guide.
What is MCP?
A standard way for AI assistants to use outside tools. Annotate's MCP lets your assistant look through a recording: the frames, your marks and what you said, at the moments that matter.
Is my screen recording private?
Yes. Recordings and transcripts stay on your computer unless you turn Cloud on. Cloud copies expire after 30 days, and your local copy is never deleted by that expiry.
What is the cloud used for?
For AI agents that run in the cloud instead of on your computer, like Claude Code on the web, Cursor cloud agents, Lovable or Figma. They can't read recordings saved on your computer, so turning Cloud on uploads the recording for them to read. It's optional, and your local copy always stays.
Is it really free?
Yes. The free plan has unlimited on-device transcription in English and needs no login. Paid plans add cloud transcription in your language, Annotate Live and cloud recordings.
Does it work on Mac and Windows?
Yes. Download it for Mac (Apple Silicon) or get it from the Microsoft Store on Windows.

Next time it's hard to explain, show it.

Free to download. No account needed to start.