Docs for Agents

A reference index of every machine-readable endpoint, format, and tool for connecting AI agents and LLMs to Kestra.

Use this page when you build an AI tool, agent, or integration that needs to read Kestra documentation or schemas. For step-by-step setup in a coding agent, start with the Docs MCP server and Agent Skills.

Docs MCP server

The Kestra Docs MCP server gives AI tools live access to plugin documentation, blueprints, and product docs. Configure it in Claude Code, Cursor, or any MCP-compatible agent.

Auto-discovery endpoint:

https://kestra.io/.well-known/mcp.json

Server URL:

https://api.kestra.io/v1/mcp

No authentication required for the public server. For per-tool configuration, see Docs MCP server. This server serves documentation only; to call your own flows from an agent, see Flow MCP servers.

Agent Skills

Agent Skills are structured knowledge files that teach AI coding agents how to generate valid Kestra flows, harden them for production, operate Kestra environments with kestractl, and migrate to Kestra 2.0.

Install all Kestra skills with:

npx skills add kestra-io/agent-skills

See Agent Skills for the list of skills, what each one covers, and example prompts.

Plain Markdown docs

Every Kestra documentation page is available as plain Markdown. Three ways to retrieve it:

• From the Copy Page menu on every docs page, choose Copy as Markdown, View as Markdown, Open in ChatGPT, or Open in Claude.

• Append .md to any kestra.io/docs/* URL:

  https://kestra.io/docs/quickstart.md

• Set the Accept header on the request:

  curl -H "Accept: text/markdown" https://kestra.io/docs/quickstart

This works for all docs pages, including versioned docs such as kestra.io/docs/1.3/quickstart.md.

llms.txt endpoints

The llms.txt file follows the llms.txt standard and includes the MCP server discovery URL.

Plugin and flow schemas

The flow schema is the same schema used by Kestra’s AI Copilot. Pass it to your agent to validate flow YAML before deploying.

OpenAPI specs

Kestra’s REST API is described by an OpenAPI specification. Use it to generate clients or give an agent a complete picture of available endpoints.

The interactive API reference at /docs/api-reference is rendered from these same spec files.