AI Integration
Purse documentation is available to AI coding assistants through two complementary mechanisms. Both are designed to help your agent understand the API and generate accurate integration code — they differ in how the agent accesses the content.
Which approach to use?
| MCP Server | LLMs.txt | |
|---|---|---|
| How it works | Live semantic search + OpenAPI access via 6 dedicated tools | Static curated overview, loaded as conversation context |
| Response quality | Precise, page-level answers with code examples | Broad structural overview of the platform |
| Tool support | MCP-compatible agents: Claude Code, Claude Desktop, Cursor, Windsurf, Gemini CLI | Any AI tool that can read a URL |
| Best for | Writing and debugging integration code | Quick orientation without tools |
| Setup time | ~2 min | ~1 min |
Recommended: start with MCP
If your AI agent supports MCP, use the MCP server. It gives your agent real-time access to the exact content it needs — including full OpenAPI spec inspection — and is purpose-built for LLM workflows.
When to use MCP
Use the MCP server when:
- Your AI agent supports the MCP protocol (Claude Code, Cursor, Windsurf, Gemini CLI, Claude Desktop, etc.)
- You're actively building a Purse integration and need precise, page-level answers
- You want to query specific API endpoints, request/response schemas, or edge cases
- You need live documentation that always reflects the latest version
When to use LLMs.txt
Use LLMs.txt when:
- Your AI agent doesn't support MCP (ChatGPT, Gemini, basic chat interfaces)
- You want a lightweight doc overview to paste into any conversation
- You need a quick platform orientation before diving into specifics
Use both together
For the best results, combine the two:
- Load LLMs.txt once to give your agent a structural overview of the platform
- Connect the MCP server for precise lookups while coding
Both are free and require no authentication.