What is MCP for social data?
MCP (Model Context Protocol) is a standard for connecting AI clients (Cursor, Claude, VS Code, and others) to tools and data sources. MCP for social data means those clients can call Social Fetch operations as tools, same catalog as the REST API, billed the same way.
How Social Fetch MCP works
Point your client at https://api.socialfetch.dev/mcp, finish OAuth, and the agent sees tools mapped from the public API. That covers the marketplace catalog across 17 platforms (137 ops), not a separate shadow schema.
You can scope the tool list to fewer platforms so the client loads less context. Responses still use the Social Fetch envelope.
MCP vs Ask
MCP exposes typed tools for each operation. Ask (POST /v1/ask) takes a natural-language question and routes to an operation, useful when the agent does not know the path yet.
Many teams use Ask to discover the route, then pin the typed MCP tool or REST path in production.
Billing and auth
Tool calls meter like the matching REST route. Credits, 402, and meta.creditsCharged behave the same. Whoami, balance, and NL Ask on MCP still require OAuth.
Install guides for Cursor, VS Code, Claude Code, and Claude Desktop live under /docs/integrations/mcp. Pair with Agent Skills so the model does not invent endpoints. Do not put a sfk_ API key in mcp.json for the hosted URL.
FAQ
Do I put an API key in mcp.json?
No for the hosted Social Fetch MCP URL. You sign in with OAuth. See /docs/integrations/mcp.
Are MCP calls billed differently?
No. Tool calls meter like the matching REST route. Credits and 402 behavior are the same.
What about Agent Skills?
Install the Social Fetch skill pack for durable instructions (read llms.txt, use Ask, never invent routes). Docs: /docs/integrations/skills.
Where is the install guide?
/docs/integrations/mcp has Cursor, VS Code, Claude Code, and Claude Desktop setup.