Passer au contenu

Build with AI

Point Claude Code, Cursor or another coding assistant at the EDFlow docs: llms.txt, llms-full.txt, a Markdown version of every page, the docs MCP server, and the rules to give your agent.

Cette page n'est pas encore offerte en français; elle est affichée en anglais.

These docs are built to be read by coding assistants as well as people.

Machine-readable docs

ResourceWhat it is
/llms.txtAn index of every page with a one-line description
/llms-full.txtEvery page in one Markdown file (under 5 MB)
Any page + .mdThe Markdown version of that page, for example /developers/quickstart.md. Requests with Accept: text/markdown get Markdown too
OpenAPI documentThe API contract the handlers validate against
Event JSON SchemasLinked from every event page

Generated tables (event types, errors, settings, roles) are rendered as real Markdown tables in these versions.

The docs MCP server

The docs expose a stateless MCP endpoint at https://docs.edflow.space/api/mcp (Streamable HTTP, JSON responses) with three tools: list_pages, get_page and search. Everything it returns is public documentation.

Claude Code:

claude mcp add --transport http docs https://docs.edflow.space/api/mcp

Cursor (.cursor/mcp.json):

{ "mcpServers": { "docs": { "url": "https://docs.edflow.space/api/mcp" } } }

Rules to give your agent

Paste these into your agent's instructions when it writes integration code:

  • Use only synthetic data. In the sandbox, identifiers start with SYN and names are synthetic.
  • Never send result values, notes, diagnoses, medications, allergies, health card numbers, addresses, phones or emails: they are prohibited fields.
  • Event ids are deterministic and never contain patient identifiers.
  • Sign requests and fetch tokens at send time, not in a transformer.
  • Retry only 5xx, 429, token_expired and re-sign errors, with backoff and Retry-After.
  • Verify webhooks with a Standard Webhooks library against the raw body; deduplicate on webhook-id; keep the highest version.
  • Never paste real messages, credentials or patient data into an AI tool.

Ask the docs

The Ask the docs button answers questions from these pages and cites them. It can't see any hospital's data. Never paste real HL7, FHIR or patient data into it; it refuses those.

Dernière mise à jour le

Cette page vous a-t-elle été utile?

Sur cette page