Record it. Your agent does the rest.

Hit record on a browser session — or drop in any video. ThinkRun hands the whole thing to your agent, which watches, understands, and proposes the fix for your review.

  1. 0:00

    Recording starts — screen, voice, clicks

  2. 0:17

    Remove npx @thinkrun/mcp from the hero

  3. 0:24

    Soften the advanced opt-in framing

  4. 0:32

    Make docs, pricing, live-example links active

  5. Applied

    3 changes shipped — on the page you are reading

The feedback that shaped this page, recorded with ThinkRun.

View this recording

Record your session.

Capture the screen, your voice, and every click. ThinkRun structures it for your agent.

Start Recording
+ Recording details

Video, transcript, clicks, and a timeline. Console and network with the extension.

Your recording is private until you share a link.

Or drop in any videoPhone recordings included · .mp4 .webm .mov

Free to start. Nothing leaves your browser until you press stop.

Record right from this site or drop in any video — nothing to install.

Works with Claude Code, Cursor, Cline, Windsurf — or paste the Markdown into any AI.

You record

Screen, voice, clicks, console, and network — captured and time-synced in one pass. From this site, the extension, or any video you upload — even a phone recording.

Your agent watches

It reads the structured recording in one pass — the events that matter, not a 200k-token page dump. No re-explaining, no guessing from pixels.

It fixes

Proposes the diff for your review — through your normal workflow, not around it. You stay the editor.

Sharing

Never get a “looks weird on my machine” again.

What you receive is a replayable session with the failing request attached — video, timeline, console, and network, in one link. Copy it as Markdown or JSON and it drops straight into an agent.

Same capture, whether the reader is your teammate — or their AI.

See a real recording
thinkrun.ai/s/9k2mQ7
MarkdownJSONFull page

## Recording: checkout 500

- 00:12 click "Pay $59.00"

- 00:12 POST /charge → 500

- 00:13 console TypeError: total undefined

“Move the CTA up and tighten the card padding.”

pinned to .checkout-cta at 00:42

Change shipped — comment resolved

Frontend feedback

Point at the screen. Skip the spec.

Walk through a page with your voice. Every comment pins to the exact element and moment, so the change ships — through your developer’s normal review, not around it.

Recording feedback? You don’t need to install anything.

Private by design

You hit record

Nothing is captured until you choose to record — and recording stops when you say so.

You review before it leaves

Passwords are masked at capture and auth headers scrubbed — and you see the full timeline before anything is shared.

Local mode

No account, no server — recordings never leave your machine unless you share them. Free forever.

And when you’re ready

Hand off your browser.

Let your agent drive the Chrome you’re already signed into — no fresh cloud browser that’s logged out of everything. Scoped to the tabs you attach, every command policy-checked, kill switch always on.

Read browser safety docs

NAVIGATE /reports

CLICK "Filter: last 30 days"

SCREENSHOT export confirmed ✓

scope: app.yoursite.com · policy: on

Three ways in.

Just record

Right from this site, or upload any video — phone recordings included. Nothing to install.

Add the extension

Console, network, and element-pinned comments come along with the video.

Connect your agent

Your agent pulls recordings itself — one-step setup in the docs.

Start recording — free