Skip to main content

Connect your AI agent via MCP

The Purse documentation exposes a Model Context Protocol (MCP) server built specifically for AI agents. It enables semantic vector search across all documentation pages and programmatic access to the full OpenAPI specifications — so your agent can find the right answer instantly instead of guessing from training data.

Built for AI-speed integration

The MCP server is optimized for Agentic workflows: tools are composable, descriptions guide the agent through the correct call sequence, and search results include page summaries so the agent can pick the best match before fetching the full content.

MCP endpoint​

https://api.purse-live.com/mcp/doc/v1/mcp

This endpoint uses the Streamable HTTP transport defined in the MCP specification.

Available tools​

The server exposes six tools organized in two groups.

Documentation tools​

ToolDescription
search_purse_docsStart here. Semantic vector search across all of docs.purse.tech — API v1/v2, Vault, Checkout Widget, SDKs. Returns 5 results, each with a URL and an AI-generated summary. Scan the summaries to pick the right page, then call fetch_purse_doc. Prefer this over list_purse_docs whenever you have a concrete question or keyword.
fetch_purse_docFetch the full markdown of a specific page (code examples, request/response schemas, edge cases). Requires an exact page URL — use search_purse_docs or list_purse_docs first to find it.
list_purse_docsBrowse the documentation tree by depth when you want to explore structure rather than search. path accepts a relative segment (e.g. integrate) or a full URL; depth=1 returns direct children only. Each result is typed as section (has sub-pages) or page (leaf). Prefer search_purse_docs when you have a concrete query.

Recommended flow:

search_purse_docs("how to create a client session")
  → returns 5 results with summaries
  → pick best match URL
  → fetch_purse_doc(url)
  → read full content including code examples

OpenAPI spec tools​

ToolDescription
list_purse_openapi_specsList all available OpenAPI spec names. Call this first when you don't know which spec to use.
list_purse_openapi_specGet a compact overview of a spec: version, servers, all method names with HTTP verbs, and all schema names. Always call this before inspect_purse_openapi_spec.
inspect_purse_openapi_specRetrieve full definitions for specific methods and/or schemas. Supports shallow=true (only directly referenced schemas) or shallow=false (full recursive schema graph). Start with shallow=true to save tokens, and use shallow=false only when you need the complete schema hierarchy.

Recommended flow:

list_purse_openapi_specs()
  → pick the right spec name
  → list_purse_openapi_spec(spec_name)
  → identify target method and schema names
  → inspect_purse_openapi_spec(spec_name, methods=[...], shallow=true)

Connection tutorials​

Claude Code​

Claude Code supports remote MCP servers via its built-in mcp add command.

Project-level setup (recommended):

claude mcp add --transport http purse-docs https://api.purse-live.com/mcp/doc/v1/mcp

User-level setup (available in all projects):

claude mcp add --transport http --scope user purse-docs https://api.purse-live.com/mcp/doc/v1/mcp

Verify the connection:

claude mcp list
Team-shared configuration

Commit a .claude/mcp.json file at your project root so the whole team gets the server automatically:

{
  "mcpServers": {
    "purse-docs": {
      "type": "http",
      "url": "https://api.purse-live.com/mcp/doc/v1/mcp"
    }
  }
}

Claude Desktop​

Claude Desktop (v0.10.0+) supports remote MCP servers natively. Edit your configuration file:

OSFile location
macOS~/Library/Application Support/Claude/claude_desktop_config.json
Windows%APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "purse-docs": {
      "url": "https://api.purse-live.com/mcp/doc/v1/mcp"
    }
  }
}

Restart Claude Desktop after saving. A hammer icon in the chat interface confirms the server is active.

Cursor​

Cursor supports MCP via a JSON config file. Create or edit .cursor/mcp.json in your project root (project-level) or ~/.cursor/mcp.json (global):

{
  "mcpServers": {
    "purse-docs": {
      "url": "https://api.purse-live.com/mcp/doc/v1/mcp"
    }
  }
}

Reload the window (Cmd+Shift+P → Reload Window) and confirm the server appears in Cursor Settings → MCP.

Read more about Cursor MCP

Windsurf​

Windsurf reads MCP configuration from ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "purse-docs": {
      "url": "https://api.purse-live.com/mcp/doc/v1/mcp"
    }
  }
}

Restart Windsurf after saving. The server appears in Windsurf Settings → MCP Servers.

Gemini CLI​

Gemini CLI reads MCP servers from ~/.gemini/settings.json:

{
  "mcpServers": {
    "purse-docs": {
      "httpUrl": "https://api.purse-live.com/mcp/doc/v1/mcp"
    }
  }
}

Start a new Gemini CLI session to pick up the change. Verify with:

gemini mcp list

Read more about Gemini CLI MCP support

Verifying the connection​

Once connected, ask your agent:

Search the Purse documentation for how to create a client session and authenticate with OAuth2.

A working MCP connection returns a live answer sourced directly from the documentation — including the relevant page URLs your agent can drill into for code examples.

Streamable HTTP requirement

All tutorials above assume your tool supports the Streamable HTTP transport. If your tool only supports SSE or stdio, use the LLMs.txt integration as an alternative.

Next steps​