⚡

STT.ai inside Claude Desktop, Cursor, and any MCP client

Transcribe audio/video, list and chat with past transcripts, summarize, analyze, and generate content — directly inside the AI tool you already use. The official Model Context Protocol server for STT.ai.

pip install sttai-mcp GitHub

✦ The only STT MCP server that lets Claude transcribe AND query in one conversation

Other transcription MCPs (Otter, Fireflies, Granola, Read AI) are read-only over stored transcripts. STT.ai's MCP server adds transcribe_url and chat_with_transcript — paste a YouTube link in Claude and immediately ask questions about it, all without leaving the conversation.

10 tools, ready to use

transcribe_url

Download + transcribe any URL — YouTube, podcast feed, direct file.

chat_with_transcript

RAG Q&A over a transcript with source citations and timestamps.

list_transcripts

Paginated list of your past transcripts, filter by status.

get_transcript

Fetch one transcript with segments, speakers, and metadata.

summarize_transcript

AI summary — cached or freshly regenerated.

analyze_transcript

Sentiment, topics, entities, action items, questions, PII.

generate_from_transcript

Blog posts, social, study guides, flashcards, podcast show notes.

export_transcript

Export as txt, srt, vtt, json, or csv.

list_models

Available transcription models with WER benchmarks.

list_languages

Languages STT.ai supports — 99 and counting.

Setup in 2 minutes

1. Install the package

pip install sttai-mcp

2. Get your API key

Grab one from your account page or sign up free for one.

3a. Claude Desktop

Edit your config file (~/Library/Application Support/Claude/claude_desktop_config.json or %APPDATA%\Claude\claude_desktop_config.json):

{
  "mcpServers": {
    "stt-ai": {
      "command": "sttai-mcp",
      "env": { "STT_API_KEY": "your-api-key-here" }
    }
  }
}

Restart Claude Desktop. The 10 STT.ai tools should appear automatically.

3b. Cursor

Settings → MCP → Add new MCP server, paste the same JSON.

Example prompts

In Claude Desktop, after configuring:

"Transcribe https://www.youtube.com/watch?v=abc123 and summarize the key points."

Claude calls transcribe_url, then summarize_transcript on the result.

"What are my last 5 transcripts about?"

Claude calls list_transcripts(limit=5) and reads each title.

"In transcript abc123, what did the speaker say about pricing?"

Claude calls chat_with_transcript and gets an answer with timestamped source citations.

Open source (MIT license). Issues, PRs, ideas welcome on GitHub.

github.com/sttaigit/sttai-mcp pypi.org/project/sttai-mcp