Serveur MCP transcript1

@transcript1/mcp

MCP (Model Context Protocol) connector for transcript1, an audio and video transcription service. Transcribe file uploads or links from YouTube, X, TikTok, Instagram, Facebook, Bluesky, Twitch, Dailymotion, Pinterest, LinkedIn and Snapchat Spotlight in Claude, Cursor, VS Code, and other MCP clients, and read the transcript with timestamps and optional speaker labels. Transcripts can be translated into other languages for free, with each translation kept as its own language and lined up with the original timestamps, so you get translated or bilingual subtitles. Links must be public, recorded videos; Sources and options lists the accepted link forms.

This package is a small stdio bridge to the hosted server at https://api.transcript1.com/mcp, for clients that launch MCP servers as commands. It wraps mcp-remote: the first time it starts, it opens transcript1 in your browser so you can sign in, or it uses an API key if you set one. Docs: transcript1.com/docs/mcp.

Connect without this package

Clients that support remote MCP servers can connect to https://api.transcript1.com/mcp directly and sign in to transcript1 in the browser. You need neither this package nor an API key.

• Claude and ChatGPT: add a custom connector with the URL https://api.transcript1.com/mcp, then sign in to transcript1 and approve access. Custom connectors depend on your plan.
• Claude Code: run claude mcp add --transport http transcript1 https://api.transcript1.com/mcp, then run /mcp and sign in.
• Cursor and VS Code: add the URL as a remote MCP server and sign in when asked. transcript1.com/docs/mcp has one-click installs.

Setup

You need a transcript1 account. Every client below runs the same command, npx -y @transcript1/mcp. Requires Node.js 18 or newer.

The first time the server starts, the package opens transcript1 in your browser. Sign in and approve access. The sign-in is saved in ~/.mcp-auth on your computer, so you do not repeat it every time.

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "transcript1": {
      "command": "npx",
      "args": ["-y", "@transcript1/mcp"]
    }
  }
}

Cursor

Add to ~/.cursor/mcp.json (or .cursor/mcp.json in a project). Cursor can also connect directly; use this package if Cursor's own sign-in fails.

{
  "mcpServers": {
    "transcript1": {
      "command": "npx",
      "args": ["-y", "@transcript1/mcp"]
    }
  }
}

VS Code

Add to .vscode/mcp.json:

{
  "servers": {
    "transcript1": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@transcript1/mcp"]
    }
  }
}

Claude Code

claude mcp add transcript1 -- npx -y @transcript1/mcp

Any other MCP client

Configure a stdio server with command npx and arguments -y @transcript1/mcp.

Use an API key instead

Browser sign-in needs a browser on the same computer as the MCP client. Over SSH, in CI or in a container, or to skip the sign-in, set TRANSCRIPT1_API_KEY to an API key (starts with t1_). Create one at transcript1.com/dashboard/account/tokens. It is shown once, so copy it.

{
  "mcpServers": {
    "transcript1": {
      "command": "npx",
      "args": ["-y", "@transcript1/mcp"],
      "env": { "TRANSCRIPT1_API_KEY": "t1_YOURKEY" }
    }
  }
}

With Claude Code:

claude mcp add transcript1 --env TRANSCRIPT1_API_KEY=t1_YOURKEY -- npx -y @transcript1/mcp

Configuration

Variable	Required	Description
TRANSCRIPT1_API_KEY	no	Your transcript1 API key. Must start with t1_. Without it, the package signs in through your browser.
TRANSCRIPT1_MCP_URL	no	Server URL. Defaults to https://api.transcript1.com/mcp. Must use https; plain http is accepted only for localhost and 127.0.0.1.

The key is passed to mcp-remote through its environment, not on the command line. Extra arguments given to the command are forwarded to mcp-remote.

Tools

• transcribe: create a transcription from a link on any supported platform and wait briefly for the result. Optional translate_to (ISO 639-1) also translates the finished transcript.
• get_transcription: status or transcript of a job, as text, SRT, VTT, or JSON, with paging for long transcripts. translation is both (default, translated line under each original), only, or none. translation_language (ISO 639-1) picks which translation to show when a transcript has several; by default the most recent one.
• translate_transcription: translate an already finished transcript into one language; free (0 minutes) and idempotent. Text only, timestamps kept.
• list_transcriptions: your recent jobs, filterable and paginated.
• create_upload: reserve a chunked upload for a local audio or video file.
• get_upload_status: see which chunks were received, to resume an interrupted upload.
• complete_upload: verify an upload and queue it for transcription.
• delete_transcription: cancel a running job or delete a finished one.
• get_account: minutes remaining, current plan, renewal date, top-up pack price, and auto top-up status.
• top_up: buy 1,000-minute top-up packs with your card on file, only when you ask. It charges right away only if this connection or API key has Can buy top-ups turned on in the dashboard (off by default) and the monthly agent limit has room; otherwise it returns a link where you can pay. Pass request_id so a retried purchase is charged once.

All tools act on your own account only.

Example prompts

Once the server is connected, ask in plain language:

• "Summarize this YouTube video with timestamps: https://www.youtube.com/watch?v=VIDEO_ID"
• "Transcribe this podcast episode with speaker labels: LINK"
• "Make SRT subtitles for this TikTok and save them as subs.srt: LINK"
• "Transcribe this video and give me an English translation with timestamps: LINK"
• "Transcribe ~/Downloads/interview.m4a" (local files need a client that can run shell commands, such as Claude Code or Cursor)
• "Give me the five best quotes from this interview, with timestamps: LINK"
• "Write meeting notes with decisions and action items from team-call.mp4"
• "How many transcription minutes do I have left?"
• "Buy one more pack of minutes"

Agent Skills

The skills folder has Agent Skills that teach an agent how to use these tools well: waiting for results, reading long transcripts page by page, uploading files in chunks, and writing from the full transcript. They work with the MCP server; they do not replace it.

Skill	Use it for
transcript1-transcribe	Transcripts of links and local files, with speaker labels and chunked uploads
transcript1-summarize	Summaries, key points, chapters, quotes and meeting notes with timestamps
transcript1-subtitles	SRT and WebVTT subtitle files

Install them with the skills CLI:

npx skills add transcript1-com/mcp

Or copy a skill folder into ~/.claude/skills/ (Claude Code, every project) or .claude/skills/ (one project). In the Claude apps, upload the skill folder as a ZIP file in the Skills settings.

Pricing

See transcript1.com/#pricing.

Support

support@transcript1.com

License

MIT. Copyright (c) 2026 transcript1.com (one1.dev).