Skip to main content

Connect your AI coding tools to Peak Gateway docs so they can search and retrieve API documentation in real time.

This endpoint is an unauthenticated, read-only search surface for the published documentation. It cannot inspect accounts, run transactions, issue refunds, or call the separately deployed Peak Gateway transaction MCP service.

Quick setup

Claude Code

claude mcp add --transport http peak-gateway-docs https://docs.peakgateway.co/mcp

Cursor

Create .cursor/mcp.json:

{
  "mcpServers": {
    "peak-gateway-docs": {
      "transport": "http",
      "url": "https://docs.peakgateway.co/mcp"
    }
  }
}

VS Code (Copilot)

Create .vscode/mcp.json:

{
  "mcpServers": {
    "peak-gateway-docs": {
      "transport": "http",
      "url": "https://docs.peakgateway.co/mcp"
    }
  }
}

Windsurf

Create ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "peak-gateway-docs": {
      "transport": "http",
      "url": "https://docs.peakgateway.co/mcp"
    }
  }
}

Other MCP clients

  • Server name: peak-gateway-docs
  • HTTP endpoint: https://docs.peakgateway.co/mcp

Most clients need these two fields:

{
  "name": "peak-gateway-docs",
  "transport": "http",
  "url": "https://docs.peakgateway.co/mcp"
}

Available tools

Search documentation.

  • Params:
    • query (string, required)
    • limit (number, 1-20, default 5)
  • Returns: titles, URLs, snippets, relevance scores

Example tool call:

{
  "name": "docs_search",
  "arguments": {
    "query": "transactions",
    "limit": 5
  }
}

docs_fetch

Retrieve full page content as markdown.

  • Params:
    • url (string, required, full URL from search results)
  • Returns: title, description, table of contents, full markdown

Example tool call:

{
  "name": "docs_fetch",
  "arguments": {
    "url": "https://docs.peakgateway.co/docs/api/list-transactions-proxy"
  }
}

Example usage (JSON-RPC)

Search for "transactions":

curl -sS https://docs.peakgateway.co/mcp \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"docs_search","arguments":{"query":"transactions","limit":5}}}'

Fetch a page returned by search:

curl -sS https://docs.peakgateway.co/mcp \
  -H 'content-type: application/json' \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"docs_fetch","arguments":{"url":"https://docs.peakgateway.co/docs/api/list-transactions-proxy"}}}'

Verify connection

  1. Run:
npx @modelcontextprotocol/inspector
  1. Add an HTTP server with URL https://docs.peakgateway.co/mcp.
  2. Call docs_search and confirm you get results.

Generated-artifact contract checks

Repository changes to the MCP plugin or Pages worker are covered by:

bazel test \
  //apps/websites/gateway/docs:mcp_artifact_contract_test \
  //apps/websites/gateway/support-docs:mcp_artifact_contract_test

These normal website CI tests depend on each site's real Docusaurus build and run the Pages worker generator and runtime against its generated mcp/docs.json. They check the document fields, initialization, tool discovery, search snippets, full-page content and heading anchors, plus malformed requests and missing pages. The separate config-only :test targets remain fast and do not build the sites.