MCP Server — FetchLayer Docs

MCP Server

FetchLayer runs a single Model Context Protocol server exposing all 37 live platforms as tools — no per-platform setup, no glue code.

Connection details

Server URL https://mcp.fetchlayer.dev
Transport streamable-http

Authentication

Use your FetchLayer API key as a Bearer token in the Authorization header — the same pattern as the REST API.

Bearer header (recommended)

Editors and agents that take a config file — Cursor, Claude Code, Codex, Gemini CLI, Windsurf, Zed — let you set a header per server. Each names the fields its own way (VS Code wants "servers" and "type": "http", Windsurf serverUrl); the setup guide in your dashboard writes the right one for you.

json
{
  "mcpServers": {
    "fetchlayer": {
      "url": "https://mcp.fetchlayer.dev",
      "headers": {
        "Authorization": "Bearer ss-your-key"
      }
    }
  }
}

Or, calling the server directly over HTTP:

bash
curl -X POST https://mcp.fetchlayer.dev \
  -H "Authorization: Bearer ss-your-key" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Inline path key (not recommended)

If your client's config has no headers field — just a URL — pass the key inline as a /k/<key> path prefix instead.

json
{
  "mcpServers": {
    "fetchlayer": {
      "url": "https://mcp.fetchlayer.dev/k/ss-your-key"
    }
  }
}

Prefer the Bearer header when you can

A key embedded in the URL can end up in logs, shell history, or proxy records. Only use the /k/ form when your client genuinely has no way to send a custom header.

Sign in instead (OAuth)

Clients that support MCP sign-in — Claude (custom connectors, on web, desktop and mobile), ChatGPT, Claude Code, Codex, Gemini CLI, VS Code, Cursor — need no key at all. Give them https://mcp.fetchlayer.dev with no header; they open a FetchLayer page where you sign in and approve, and they keep the connection. In ChatGPT choose OAuth, or No authentication with a /k/ address — not Mixed.

Tip

Don't have a key yet? Sign up for a free API key — no credit card required.

Fewer tools: ?platforms=

A full-access key lists 243 tools. Some clients send only so many: Windsurf loads at most 100 across all its servers, VS Code and the OpenAI API (Responses, Agents SDK) at most 128 per request, and ChatGPT picks tools less reliably past about 70. Add ?platforms= to any address and the connection sees only those platforms:

json
{
  "mcpServers": {
    "fetchlayer": {
      "url": "https://mcp.fetchlayer.dev/mcp?platforms=reddit,tiktok",
      "headers": {
        "Authorization": "Bearer ss-your-key"
      }
    }
  }
}

It works on every form of the address, including https://mcp.fetchlayer.dev/k/ss-your-key?platforms=reddit, and as an X-FetchLayer-Platforms: reddit,tiktok header. The ids are the ones in the list below. It only ever narrows: a key restricted to some platforms in the dashboard never sees more than those, whatever the address asks for. An id we don't know is refused with an error that names it.

Tools

Every endpoint is a tool named <platform>_<endpoint> — for example tiktok_user_profile. The first platforms we shipped — Reddit, X/Twitter, YouTube, Google Maps and the Play Store — keep their original scrape_ names (scrape_search for Reddit, scrape_twitter_search). Call tools/list for the full, always-current list.

Beside each platform: its id for ?platforms=, and how many tools it contributes. Docs for each are one click away.