Skip to main content

​
You own the code

Generated code is yours to modify. There’s no lock-in, no proprietary runtime, no SDK dependency. The server is standard Python using open-source libraries (FastMCP, httpx, Pydantic).

​
Common customizations

​
Adjust tool descriptions

Tool descriptions are used by AI agents to decide when to call a tool. You can improve them directly in server.py: 
@mcp.tool()
async def list_users(page: int = 1, limit: int = 10) -> dict[str, Any]:
    """List all active users with pagination.

    Use this to browse the user directory. Returns name, email, and role
    for each user. Supports pagination with page and limit parameters.
    """
Better descriptions lead to better tool selection by AI agents.

​
Add custom tools

Add new tools directly in server.py: 
@mcp.tool()
async def my_custom_tool(query: str) -> str:
    """A custom tool that combines data from multiple endpoints."""
    users = await list_users(limit=5)
    # ... your custom logic
    return result

​
Modify resilience settings

All resilience parameters are in .env and can be changed without modifying code. See Security Features for all options.

​
Change the base URL

Override the API base URL in .env:
.env
BASE_URL=https://staging-api.example.com
Useful for switching between production, staging, and local development environments.

​
Re-generating

If the original API spec is updated, you can re-generate the server from the updated spec. Your customizations in server.py will be overwritten, so keep custom tools in a separate file if you plan to re-generate frequently.
A good pattern is to keep customizations in a wrapper file that imports from the generated server, so re-generation doesn’t overwrite your changes.