Post Reef

npx add-mcp 'https://postreef.com/mcp'

MCP server

Post Reef runs a remote Model Context Protocol server, so an AI assistant (Claude, ChatGPT, Cursor, VS Code, or any MCP client) can quote and run extractions, and read the results, on your account. It runs the same pipeline as the REST API: same prices, same cache, same credits.

Server URL

https://postreef.com/mcp

Transport: Streamable HTTP.

Connect

Clients that support OAuth (Claude, ChatGPT, Cursor, VS Code) only need the URL: the first time a tool runs, your browser opens Post Reef, you sign in and click Allow. To use an API key instead, create one at https://postreef.com/developers/api-keys and send it as Authorization: Bearer pr_… (or x-api-key: pr_…).

Claude Code

# OAuth: sign in from the browser on first use
claude mcp add --transport http postreef https://postreef.com/mcp

# Or with an API key
claude mcp add --transport http postreef https://postreef.com/mcp --header "Authorization: Bearer pr_..."

Claude (claude.ai and desktop)

Settings → Connectors → Add custom connector, name it Post Reef and paste https://postreef.com/mcp.

Cursor

~/.cursor/mcp.json:

{
  "mcpServers": {
    "postreef": {
      "url": "https://postreef.com/mcp",
      "headers": { "Authorization": "Bearer pr_..." }
    }
  }
}

Leave out headers to sign in with OAuth instead.

VS Code

.vscode/mcp.json:

{
  "servers": {
    "postreef": {
      "type": "http",
      "url": "https://postreef.com/mcp",
      "headers": { "Authorization": "Bearer pr_..." }
    }
  }
}

ChatGPT

Settings → Apps → Create app (developer mode), paste https://postreef.com/mcp and choose OAuth. Results show as an extraction card in the chat.

Tools

• quote. What a link is (title, duration, subtitles, comments) and the exact credit cost of an extraction with the given parameters. Free; starts nothing.
• extract. Starts an extraction and waits up to 50 seconds for it. Without a schema it is a download (parts: transcript, comments, audio, video); with schemaId (e.g. Recipe) or an inline schema it also returns structured JSON (inputs picks what the AI reads). Debits credits; refunded in full if it fails. Identical runs from the last 30 days are free.
• get_extraction. Status and, once complete, metadata, transcript, comments, the extracted JSON and links to every file. Can wait up to 50 seconds for a running extraction.
• list_extractions. Your recent extractions, newest first.
• list_schemas. The predefined schemas and your own, optionally with one schema's full definition.
• get_profile. The connected account and its credit balance.

File links in results are the same /v1/extractions/{id}/files/{name} URLs the API returns: they open with your API key or in a browser where you are signed in.

Limits

MCP calls share the API's per-minute rate limits and your account's concurrency limit. See Errors & limits.