The AI agent-native screen recorder
Screen Recordings for AI Agents
Your agents read your recordings, and now record their own. Record once and a teammate gets the video while Claude Code, Cursor, or Codex get the summary, the exact moments you pointed at with frames and click targets, and the transcript. Or let the agent do the recording: it captures the app it just built, headless or on your real screen, and reads it straight back.
New to the idea? Start with what is an agent-readable recording.
Free plan, no card. Clipy Pro is $9/mo, cancel anytime.
One recording. Real work, done.
The same Clipy link a teammate watches is the context your agent acts on. Here is what that looks like across the team.
Repro the bug once, hand the agent a real reproduction
What used to be a written repro with numbered steps and cropped screenshots is now a single recorded run.
- Every click carries its element target, so the developer gets a repro that actually reproduces.
- Paste the link into your coding agent and it reads the frames, not a paraphrase of them.
- The agent can generate the Playwright spec straight from the recorded flow.
One demo becomes the whole launch kit
What used to be a morning of writing the changelog, the help doc, and the social thread by hand starts from one recorded demo.
- Demo the feature once and let the agent draft the changelog, help article, and social copy.
- The recording's frames drop in as screenshots, so nothing is described from memory.
- Every asset stays anchored to what you actually showed on screen.
Talk through the change, ship the change
What used to be a 20-minute writeup of what you want built is now a two-minute walkthrough your agent can act on.
- Point at the UI, say what should change, and share one link.
- Your coding agent reads the summary and the pointed-at frames, then opens the fix.
- No ticket-writing tax between the idea and the pull request.
Record the walkthrough, let the agent do the follow-up
What used to be a recap email typed from scratch after every call starts from the recording itself.
- Send a personalized async demo instead of scheduling another live call.
- The transcript and summary become the follow-up email and the CRM note.
- Prospects rewatch the exact moment that mattered, on their own time.
Walk the flow once, get the help article for free
What used to be a support reply written and re-written by hand is generated from the recorded flow.
- Do the flow once and let the agent turn it into a numbered help article with a screenshot per step.
- Or hand engineering a clean escalation they can reproduce, not a vague description.
- The same recording answers the next customer who hits the issue.
Give Claude Code, Cursor, or Codex a bug it can see
What used to be "it breaks when I click here" over three Slack messages becomes one link your agent reads end to end.
- Record “when I click THIS, it breaks” and paste the link into your agent.
- It reads the frames, greps the codebase for the button it can literally see, and opens the fix.
- Click coordinates resolve “this” to real pixels, even with three Export buttons on screen.
One command wires Clipy into your agent
No API key to copy or paste. Run this, approve once in the browser, and your coding agent can record and read recordings.
npx @clipy/cli agents install claude # also: codex · cursor
- Step 1
The browser opens; click Approve once. It names the client and the scopes: read your recordings, and record and upload.
- Step 2
The Clipy skill installs into your agent's skills directory.
- Step 3
Your agent records with Clipy and reads any recording back.
Full setup, the manual paste-a-key fallback, and every command live in the CLI docs.
Your agent records its own runs
Reading a recording is half of it. With the Clipy CLI (or the MCP server for the first two) a coding agent records the app it just built and reads the result straight back, three ways, no human holding the camera.
Headless capture
clipy record --url opens the app in a headless browser and uploads the run, with no display, so it works in CI and cloud sandboxes.
Live session, narrated
clipy session start records in the background while the agent drives the app with its own tools. Each clipy mark becomes a transcript chapter, labeled as agent narration, never passed off as speech-to-text.
The real Mac screen
--source mac-screen records your actual logged-in screen (a chosen window or display) through the running Clipy Mac app, with the recording indicator always visible.
Every mode returns a normal Clipy link with a transcript, summary, and key moments. Full commands in the CLI docs and MCP docs.
Why an agent can act on a Clipy recording
Today's agents don't watch video. Clipy does the translation server-side, so a plain link works where a raw MP4 is a dead end.
Just a URL: add .md
Append .md to any public watch link and it serves a context document: the summary, key moments, frames, and transcript. No install, no API key, and far fewer tokens than an MCP round-trip. The low-friction, plain-text access agents increasingly prefer. Humans paste the normal link; agents with the skill know the convention.
Key moments, not just words
Clipy finds every spoken pointer and extracts the frame at that instant. When a click track exists (Mac app and Chrome-extension tab recordings) it fuses the real click coordinates, and on Mac app recordings, the clicked element's role and label. So a moment reads “clicked the Settings button at (0.63, 0.41)”, a repro that actually reproduces, not a paraphrase.
MCP for private libraries
Use the skill for pasted links. Use MCP when the agent needs authenticated access: searching your whole library, reading private recordings, or pulling key-moment frames inline through a tool call.
https://clipy.online/video/<id>.md
# UI Recording Feedback
Screen recording · 1:13 · watch: https://clipy.online/video/cey8aix0…
## Summary
The speaker reviews a recording interface, pointing out a duplicated
timer and a settings section that should be collapsed…
## Key moments
```json
[{"t_ms":6400,"caption":"points to 'this section' on screen",
"frame_url":"https://cdn.clipy.online/key-moments/…/6400.jpg",
"x":0.63,"y":0.41,"source":"fused","confidence":0.9}, …]
```
### 0:06 · points to 'this section' on screen
## Transcript
```
[0:01] Hey great. So there are a few things I'd like you to update.
[0:06] For example this section…
```The MCP tools your agent gets
npx -y @clipy/mcp # CLIPY_API_KEY via: npx @clipy/cli login
get_agent_context
One-call bundle: metadata + summary + key moments (frames inline as images) + transcript. Start here.
get_key_moments
The visual pointers: what the speaker pointed at, the frame at that instant, click coordinates when a click track exists (Mac app + Chrome-extension tab recordings).
get_transcript
Full timestamped transcript (segments + plaintext).
get_summary
AI summary: TL;DR, key points, action items.
search_recordings
Find recordings by keyword.
Full reference at clipy.online/docs/mcp · install the skill.
Plug your recordings into your workflow
Every recording becomes something your agents can act on: a timestamped transcript, chapters, and a markdown twin Claude Code, Cursor, and Codex read straight from the share link. Start free, upgrade when you outgrow it — $9/mo, cancel anytime.
- Unlimited recordings, in 4K
- MCP for Claude, Cursor, Codex
- Agent-readable .md on every link
- Transcripts, summaries & chapters
- Ask-the-video Q&A
- Search across your whole library
- YouTube & video context import
- Unlimited Loom imports
The free plan keeps recording, sharing, and transcripts — no card, no expiry date.
Common questions
How do I let Claude or Cursor read a screen recording?
Default path: install the Clipy skill once, then paste the normal Clipy video link. The skill fetches the .arec context document, waits if needed, and reads the requests, plan, key moments, diagnostics, gaps, and transcript. Advanced path: install MCP for private recordings and library search.
Does this work with private recordings?
Public links work through the skill without an API key. Private and restricted recordings need credentials. Use the Clipy MCP server or CLI when an agent needs authenticated access to your private library.
Which agents and tools are supported?
The skill is the default path for Claude, Codex-style skill folders, and skill-aware agents. The underlying .arec link works anywhere a tool can fetch a URL. MCP remains available for Claude Code, Claude Desktop, Cursor, Cline, Windsurf, and other MCP clients.
What exactly are “key moments”?
Timestamped instants where the speaker referenced something visible: “this button”, “this error”, “watch what happens”. For each one, Clipy extracts the video frame at that moment; on Mac app recordings (and Chrome-extension tab recordings) it also fuses the real click coordinates. It's how an agent resolves “this” to actual pixels.
Can my agent record on its own, not just read?
Yes. With the Clipy CLI or MCP server, an agent can record a web app headlessly (built for CI and cloud sandboxes) or run a live session while it drives the app with its own tools. The Clipy CLI can additionally record your real Mac screen through the running Clipy app. The recording comes back as a normal Clipy link with a transcript, summary, and key moments: the same thing it would read from anyone else. Recording needs an ingest-scoped key; reading public links needs no key at all.
Is there a free trial?
There is no separate trial: the free plan is the trial. Record, share, and read your recordings with agents without entering a card, for as long as you want. Upgrade to Pro when you need more.
What is free and what needs Pro?
The recorder, share links, transcripts, summaries, key moments, the .md documents, and the MCP server are all usable for free, with no signup wall for viewers and no cap on how long a recording runs. The free plan covers your first 15 recordings or 2 hours of recording, whichever comes first — a one-time allowance with no expiry date, which deleting recordings does not restore. Pro ($9/mo, or $6/mo billed yearly) lifts that cap and adds 4K, whole-library Loom migration, end-of-video CTAs, and priority support.
Is my recording used to train AI models?
No. Recordings are processed to produce your transcript, summary, and key moments, and that's it. See the Clipy Pledge at clipy.online/pledge.
The workflow this unlocks
- 1. Anyone records. A PM, a QA, a customer: “when I click THIS, it breaks”. No ticket-writing, no screenshots, no repro steps.
- 2. One link travels. The same URL works in Slack for humans and in an agent's chat when the Clipy skill is installed.
- 3. The agent acts. It reads the summary, looks at the pointed-at frames, greps the codebase for the button it can literally see, and ships the fix.
Related reading
Deeper guides on handing a recording to your AI agent.