Docs for AI agents | Arbol AI Docs

These docs are published for AI tools as well as for people. An agent can read any page as Markdown, find every API reference and its OpenAPI document from a few well-known URLs, and install a skill that teaches it to call the Arbol API. All of it is public and needs no credential.

Start at llms.txt

URL	What it holds
`/llms.txt`	The index of the docs: the top-level pages and an index per section.
`/<section>/llms.txt`	One section’s pages, linked to their Markdown, such as `/en/developers/guides/llms.txt`.
`/api/<service>/llms.txt`	One API reference, such as `/api/raiz/llms.txt`, with an index per resource, such as `/api/raiz/agents/llms.txt`, that lists each operation and what it does.
`/llms-full.txt`	Every page of the docs in one Markdown document, for tools that read everything at once.

Markdown for every page

Add index.md to a page’s URL to get it as Markdown, with components turned into plain text:

https://docs.getarbol.com/en/developers/guides/create-an-agent/
https://docs.getarbol.com/en/developers/guides/create-an-agent/index.md

Docs pages also serve their authored source at index.mdx; API reference pages have only index.md. The home page is /index.md. Each HTML page links its Markdown version with <link rel="alternate" type="text/markdown">, describes itself with JSON-LD, and carries a visually hidden note that points agents to its Markdown and to /llms.txt.

OpenAPI documents

Each API reference publishes its OpenAPI document as one self-contained JSON file, with every operation’s code samples (x-codeSamples) and, in its description, the scope it needs, the credentials it accepts, its effect and its limits:

https://docs.getarbol.com/api/<service>/openapi.json

Discovery

URL	What it holds
`/.well-known/api-catalog`	An RFC 9727 API catalog: one entry per API reference, linking its page, its Markdown and its OpenAPI document.
`/.well-known/ard.json`, `/.well-known/ai-catalog.json`	The same discovery manifest: `/llms.txt`, the home page’s Markdown, each OpenAPI document and the Agent Skills index.
`/.well-known/agent-skills/index.json`	The Agent Skills this site publishes, each with a SHA-256 digest.
`/robots.txt`, `/sitemap-index.xml`	The crawler policy and the sitemap.

The Arbol API skill

/.well-known/agent-skills/arbol-api/SKILL.md is an Agent Skill that teaches an agent to call the Arbol API correctly: the base URL, how to find a service and its OpenAPI document, credentials and the Authorization header, the response envelope and errors, when not to retry, and least privilege. Install it in any tool that supports Agent Skills, or point the agent at its URL.

Skills

This site also publishes skills for specific tasks, each at /.well-known/agent-skills/<name>/SKILL.md. The index, /.well-known/agent-skills/index.json, lists them all.

Skill	What it’s for
`arbol-api`	Call the Arbol API correctly: credentials, responses, errors and retries (above).
`arbol-docs-help`	Help a clinic’s staff member do a task in Arbol’s apps from these docs’ guides.
`raiz-agents`	Create, test and update a Raíz agent, and put it on a channel.
`raiz-instructions`	Write or improve a Raíz agent’s instructions, with clinical safety limits and test conversations.
`raiz-conversations`	Follow up conversations and contacts, take over or hand back a conversation, and summarize it for a person.
`agenda-scheduling`	Find a time and book, reschedule or cancel an appointment with the person’s confirmation.
`agenda-setup`	Read Agenda’s services, agendas, sites and hours, check that a service produces slots, and guide the administrator through the app.
`arbol-troubleshooting`	Diagnose a problem, read the error, decide whether to retry, fix or escalate, and report it without patient data or credentials.
`arbol-cli`	Use the `arbol` CLI from a terminal, a script or CI without exposing credentials.
`arbi`	Work alongside Arbi, the team’s assistant: what it can do in each app and on WhatsApp, what waits for approval, who sees its notes, and when to send a person to Arbi, a guide or the Arbol team.

Search from the browser

In browsers that support WebMCP, each page registers a search_documentation tool that searches these docs and answers titles, anchored URLs and excerpts:

{ "query": "create an agent", "limit": 5 }

query is required, up to 500 characters; limit is 1 to 20, 5 by default.

Act on Arbol

These docs explain the API; they don’t call it. To let an AI assistant work with your organization’s data, such as agents, conversations or appointments, connect it to Arbol’s MCP server, which offers API operations as tools and runs them with your own credentials. See MCP.