framefetch-mcp

Local stdio MCP bridge for FrameFetch — turn any social-video URL into metadata, insights, a Whisper transcript, parametric frames, and the on-screen text burned into those frames (OCR). Works across YouTube (incl. Shorts), TikTok, Reddit, Instagram and Pinterest.

npx framefetch-mcp

Add it to any MCP client:

{
  "mcpServers": {
    "framefetch": {
      "command": "npx",
      "args": ["-y", "framefetch-mcp"],
      "env": { "FRAMEFETCH_API_KEY": "<key>" }
    }
  }
}

Tools

| Tool | What it does | Billed | |---|---|---| | framefetch_extract | One video URL → metadata, insights, transcript, frames, OCR text, digest, structured data, comment sentiment | yes | | framefetch_search | Find videos by query, returns real result URLs | yes | | framefetch_platform_capabilities | Which fields each platform supports right now | free | | framefetch_account | Your balance and recent spend | free |

tools/list works without a key. Tool calls need either FRAMEFETCH_API_KEY or an x402 payment.

Getting a key

No card, no signup form — one call returns a working key with free starting credit:

curl -X POST https://framefetch.net/v1/keys \
  -H 'content-type: application/json' \
  -d '{"email":"[email protected]"}'

An agent can do this itself and retry immediately with Authorization: Bearer <key>.

Configuration

| Env var | Default | Purpose | |---|---|---| | FRAMEFETCH_API_KEY | — | Bearer key for authenticated calls | | FRAMEFETCH_MCP_URL | https://framefetch.net/mcp | Override the remote endpoint |

The bridge forwards tools/call to the hosted server and serves tools/list from the remote too, falling back to an embedded snapshot when the network is unavailable — so introspection never fails.

Prefer HTTP over stdio? Point your client straight at https://framefetch.net/mcp. Prefer a JS/TS library? Use framefetch.

Source: framefetch-client · Docs: framefetch.net/docs

MIT