OpenAI WebMCP Challenge entry - the agent tools on this page are real. Watch the rail.
transcriptly

GitHub

Any video URL,
one clean transcript

Paste a link or drop a file. Open source, four ways in: a CLI, an MCP server, a REST API, and WebMCP tools that browser agents call right on this page.

or just try it below - no install
Try:
Agent panelnot detected

Checking for WebMCP…

  • get_transcript

    url → transcript

  • get_video_info

    url → title, duration, captions

  • transcribe_file

    the file you dropped → transcript

Activity
  • No agent calls yet.

Use it wherever your work happens

CLI

One command, fully local.

Whisper runs on your machine. YouTube, TikTok, local files.

MCP

Give your agent ears.

get_transcript + get_video_info in Claude, ChatGPT, anywhere MCP works.

REST

This site's demo API.

Demo only: 5/hour, 30-minute cap. Self-host it for real workloads.

WebMCP

Tools on this very page.

Browser agents call get_transcript natively. Watch it in the rail above.

5 transcriptions an hour, 30 minutes a video

The CLI has no limits and runs entirely on your machine. Want longer videos on hosted compute instead? Leave your email and tell me it's worth building.