We extract as much information as possible from every clip. Anything that crosses your screen becomes available to your agent. Everything’s searchable and actionable.
ClipCabinet turns your clip into a transcript, captioned frames, and every URL, error, code snippet, and ticket ID that crossed the screen, all searchable and readable by your agent. Ask it “what was the error at 2:14?” and it quotes the actual error.
The recording extension quickly records your active tab, system, and mic audio. No fuss, no setup, just click, talk, and go.
The moment you hit stop, ClipCabinet starts turning your clip into a fully indexed record your agent can query.
We have all the tools so your agent can work autonomously, with instructions generated automatically on the fly. ClipCabinet speaks Model Context Protocol natively, so Claude Code, Cursor, and Claude Desktop already know how to query it. Paste the config, mint a token, and your agent does the rest.
list_recordingsRecent clips by date or kind (video / screenshot).
search_recordingsSemantic search across the whole library.
get_recordingFull details: transcript, frames, artifacts, signed media URLs.
search_within_recordingFind moments inside one clip.
get_frameA captioned keyframe at a given timestamp.
get_audio_clipAn audio slice between start and end timestamps.
list_artifactsEvery URL, error, or ticket ID across clips.
{
"mcpServers": {
"clipcabinet": {
"type": "http",
"url": "https://mcp.clipcabinet.com/mcp",
"headers": {
"Authorization": "Bearer kg_xxxxxxxxxxxxxxxxxxxx"
}
}
}
}Every clip is private until you decide to share it. Share links are public, opt-in, and revocable. Nothing is locked in. Export any clip as markdown and send it where it's needed.
Comment at a pixel, not a timestamp.
Almost every tool out there only allows comments that are time-based. ClipCabinet's comments live on an X-Y axis, not just a timestamp. Click anywhere on a paused frame or a screenshot to drop a pin at that exact spot.