Build with AI
Build with AI

Build with AI

Feed an agent the whole API in one shot: llms.txt, llms-full.txt, openapi.json, and a Markdown twin of every page.

View as MarkdownUpdated Aug 19, 2026

These docs are published in machine formats alongside the HTML, so an assistant can ingest the whole API without scraping:

FileWhat it is
/developers/llms.txtThe documentation index — every page with a one-line description and its Markdown URL. Start here.
/developers/llms-full.txtEvery page, full text, one file. For models with a large context window.
/developers/openapi.jsonThe OpenAPI 3.1 contract — every operation, schema, error and code sample. Feed it to a client generator, Postman, or an agent.
Any docs URL + .mdEvery page has a Markdown twin — append .md to its URL. This page is at /developers/build-with-ai/overview.md.

Responses from /developers pages also carry Link discovery headers pointing at llms.txt, llms-full.txt and openapi.json, so tooling can find them without prior knowledge.

Recipes#

  • Claude / ChatGPT — paste the agent prompt; it tells the model to fetch openapi.json itself, so it stays correct as the API grows. Keep the key in the tool's environment or a connector secret, not in the chat.
  • Claude Code / other coding agents — point it at https://www.dropfans.io/developers/llms.txt and let it pull the .md pages it needs; keep DROPFANS_API_KEY in the shell environment.
  • Cursor — add https://www.dropfans.io/developers/llms-full.txt as docs (@docs), then reference it while writing your integration.
  • Codegenopenapi.json validates as OpenAPI 3.1 and carries x-codeSamples, so generated clients and rendered references both come out usable.
Never paste an API key into a shared or public chat

A key is full account access for that creator. Use an assistant that keeps secrets in its own environment, and rotate any key that ever lands in a transcript.

No MCP server yet — when one exists it will be announced in the changelog.

Next: Agent prompt or API reference.

Questions? [email protected]