Use with AI
Coding assistants guess at APIs they haven’t seen. Hand them the docs up front and they write requests that actually work.
llms.txt
Section titled “llms.txt”Plaza publishes two plain-text summaries of the API, written for language models:
| File | What’s in it |
|---|---|
plaza.fyi/llms.txt | A short overview: what the API does, how to authenticate, and the main endpoint groups |
plaza.fyi/llms-full.txt | The longer version, with more detail on each endpoint |
Start with llms.txt. If the model keeps getting parameters wrong, switch to llms-full.txt.
To use them, paste the URL (or the file contents) into your chat, or add it to your agent’s context. In Cursor, for example, you can add it under Docs. In Claude Code, mention the URL in your prompt or save it into your project’s CLAUDE.md.
OpenAPI spec
Section titled “OpenAPI spec”For tools that read OpenAPI directly, the full spec is at plaza.fyi/api/openapi. It’s the same document behind the API Reference.
MCP server
Section titled “MCP server”An MCP server for Plaza is on the way. It will let MCP-capable assistants call Plaza endpoints as tools without you writing any glue code. Until then, llms.txt and the OpenAPI spec are the best way to give an agent access.
- Tell the model that coordinates are
[longitude, latitude]. That’s the GeoJSON order, and models flip it often. - Keep your API key out of prompts. Have the agent read it from an environment variable such as
PLAZA_API_KEYand send it in thex-api-keyheader. - For complex spatial queries, ask the model to write PlazaQL instead of chaining several REST calls.