Getting Started

Clip Cabinet turns a quick screen recording into something you can actually search, share, and hand to an AI agent. You record a tab, Clip Cabinet transcribes the audio, captures the key frames, pulls out the URLs and errors and code you ran into, and writes a title and summary. The result is a clip you can find by what happened in it, not by guessing the filename.

This guide walks you through the core flow: install the recorder, capture a clip, share it, and connect your clip library to an AI agent over MCP.

What you can do with Clip Cabinet

Record a browser tab or grab a screenshot straight from a Chrome extension.
Get an AI-searchable clip: every recording is transcribed, key-framed, captioned, and summarized automatically.
Search your library by transcript text, summary, or title from the dashboard.
Share a clip with a public link, or request timestamped feedback from a reviewer.
Export any clip as Markdown and query your library from an AI agent over MCP.

Install the recorder

Recording happens through the Clip Cabinet Chrome extension.

Open the Install page and download the extension package.
Unzip it and keep the folder somewhere stable. Chrome reads from it every time it starts.
Go to chrome://extensions, turn on Developer mode (top-right), and click Load unpacked.
Point it at the unzipped folder. It shows up as ClipCabinet recorder; pin it to your toolbar.
Connect it to your account: open the extension's Settings, then paste an API token from /settings/tokens (mint a new one if you don't have one) and save.

The extension talks to https://clipcabinet.com in production.

Capture a clip

Click the extension icon on any tab. You get two ways to capture:

Record opens a recorder window and starts capturing the tab. Click stop when you're done. Use this for walkthroughs, bug repros, and anything with audio.
Screenshot opens three capture modes: Capture Visible Page (what's on screen right now), Capture Part of the Page (drag to select an area), and Capture the Full Page (scroll and stitch the entire page, including everything below the fold). The image uploads immediately and opens in its own clip. See Capturing screenshots for the full walkthrough.

Screenshots can't be taken on Chrome's own pages (chrome://), extension pages, or the Web Store. Use Record on those instead.

What happens after you stop

Video clips run through an automatic pipeline, and the detail page updates as each stage lands:

The video is finalized and the player appears.
Audio is extracted and transcribed into a timestamped transcript.
Key frames are pulled from scene changes and captioned.
Artifacts are extracted: URLs, errors, code, file paths, and ticket references mentioned in the clip.
A title and summary are written, and the clip is indexed for search.

You don't have to wait for all of it. The video plays as soon as it's ready, and the transcript, frames, and artifacts fill in as they finish.

Find a clip later

Your Dashboard lists your recent clips with thumbnails, summaries, source domain, and tags. From there you can:

Search across clip titles, summaries, and transcript text, and a match jumps you to the exact transcript line.
Filter by tag (tags are case-insensitive).
Star clips you want to keep close.

Open any clip to watch it alongside its transcript, captioned frames, extracted artifacts, and comments.

Share a clip

On a clip's detail page, turn on sharing to get a public link at /s/<token>. Anyone with the link can watch the clip and read its transcript and summary, no account needed. The link unfurls with a preview image, title, and summary. Inline video playback works in Discord, X, and iMessage; Slack shows the preview image. You can disable the share link at any time to revoke access.

Request feedback

Want comments instead of just views? Create a feedback request from the clip. That generates a separate link where a reviewer can leave comments pinned to a moment in the clip anchored to a timestamp and a spot on the frame. Close the request to stop new comments; the ones already left stay on your clip.

Export and use with AI agents

Every clip is available as Markdown. Open View as Markdown on a shared clip to get the transcript, frames, and artifacts as plain text you can paste anywhere.

To let an AI agent work with your whole library, connect Clip Cabinet over MCP. Mint a token at /settings/tokens, paste the config into your MCP client (Claude Code, Cursor, Claude Desktop, or any MCP-aware client), and your agent gets seven tools:

list_recordings: your recent clips, newest first.
search_recordings: find clips across your library.
get_recording: full detail for one clip.
search_within_recording: find a moment inside a clip.
get_frame: a captioned frame image.
get_audio_clip: a slice of the audio.
list_artifacts: the URLs, errors, and code pulled from a clip.

Authentication is a bearer token in the Authorization header; tokens are hashed at rest and can be revoked from /settings/tokens.

Plans

The Free plan includes 20 clips total, recordings up to 5 minutes each, and 30-day retention. Pro lifts the cap to unlimited clips, allows recordings up to 15 minutes, keeps clips indefinitely, and unlocks outgoing webhooks. No card is needed to start.

Troubleshooting

Chrome shows a "Developer mode" banner on chrome://extensions. This is expected while the extension is loaded unpacked and is harmless.
A "may have been corrupted" warning can appear after restarting Chrome with an unpacked extension. Dismiss it.
Nothing happens when you click Screenshot. You're probably on a chrome://, extension, or Web Store page, where Chrome blocks the capture overlay. Switch to a normal site, or use Record instead.
The extension can't reach your account. Re-check the API token in the extension's Settings and mint a fresh one at /settings/tokens if needed.
You've hit "20 / 20 free clips". You're at the Free plan's lifetime cap. Upgrade to Pro from the extension or /settings/billing to keep recording.