---
title: SkillHeart developer resources
description: The MCP server, OpenAPI description and every machine-readable file SkillHeart publishes.
canonical: https://skillheart.org/en/developers
last-updated: 2026-08-22
---
# SkillHeart developer resources

Every machine-readable endpoint SkillHeart publishes, at a stable URL. Public, read-only and free of credentials: there is nothing to request and nothing to sign up for.

## SkillHeart MCP server

A read-only Model Context Protocol server over Streamable HTTP. No API key, no OAuth, no rate limit to negotiate: POST a JSON-RPC 2.0 message and it answers. It publishes 3 tools (get_organization_info, get_programmes, list_pages) and the home page of each language as a readable resource.

### SkillHeart MCP endpoint

- URL: https://skillheart.org/mcp
- Format: JSON-RPC 2.0
- Streamable HTTP transport. Send initialize, then tools/list, tools/call, resources/list or resources/read.

### SkillHeart MCP server card

- URL: https://skillheart.org/mcp/server-card
- Format: JSON
- Name, version and remote transports, in the official MCP server-card schema. Also served at the well-known alias.

### SkillHeart MCP registry manifest

- URL: https://skillheart.org/server.json
- Format: JSON
- The same server described with the MCP registry schema, for clients and registries that look for server.json.

## SkillHeart machine-readable files

Everything a client needs to describe this surface without reading a single page of HTML. All of them are public, cacheable and served with the content type their specification defines.

### SkillHeart OpenAPI description

- URL: https://skillheart.org/openapi.json
- Format: OpenAPI 3.1
- The contract of the read-only HTTP surface: the MCP endpoint, every discovery document and the health probe.

### SkillHeart llms.txt

- URL: https://skillheart.org/llms.txt
- Format: Markdown
- What this association is, when to use it, what it does not do, and where each thing lives. Written for a model, not for a crawler.

### SkillHeart AI catalog

- URL: https://skillheart.org/.well-known/ai-catalog.json
- Format: JSON
- One hop from the domain to every agent-facing document, so nothing has to be guessed.

### SkillHeart API catalog

- URL: https://skillheart.org/.well-known/api-catalog
- Format: Linkset
- RFC 9727 linkset pointing from the MCP endpoint to its OpenAPI description and to this page.

### SkillHeart sitemap

- URL: https://skillheart.org/sitemap.xml
- Format: XML
- Every URL that exists, in four languages, with its alternates.

### SkillHeart robots and content signals

- URL: https://skillheart.org/robots.txt
- Format: Text
- Crawling rules and the content-usage preferences of the association, declared with Content-Signal.

## How SkillHeart behaves with a client that is not a browser

### Ask for markdown and you get markdown

Send `Accept: text/markdown` to the home page of any language and the response is markdown, with `Vary: Accept` so no shared cache mixes the two representations. The MCP server exposes the same text as a resource, one per language.

### A 404 that tells you where to go

A path that does not exist answers with a real HTTP 404 on the first request, not on the second, and clients without a browser get a short markdown body: what was missing, and the links to llms.txt, the sitemap and the main pages. Browsers get the designed page instead.

### Terms, and what this is not

Everything here is public and read-only. There are no credentials to request, no write operations and no webhooks, so nothing you send can change anything. The facts these endpoints return are the same ones the website publishes; when they disagree, the website is wrong and admin@skillheart.org is where to say so.