Working with AI agents
When to use MCP, Markdown, or a Skill, and what information you give the AI.
Coding agents can look up how to use genko.me too. There are three ways to connect: MCP, Markdown, and a Skill.
The public documentation is readable without an account. After you sign in with OAuth, MCP lets you read and edit entries in your own workspace as well as the product documentation.
Choose by goal
| Goal | Method | Read next |
|---|---|---|
| Let an AI read the public spec | Markdown, llms.txt | Read as Markdown |
| Manage workspace entries with an AI | OAuth MCP | MCP server |
| Pass page content to a prompt or a document | Markdown, llms.txt | Read as Markdown |
| Hand over the working rules your project follows | Skill | Skill |
You can use MCP and a Skill together. The Skill teaches the basic steps, and the agent looks up fine details through MCP. MCP requires you to sign in with a genko.me account, and you don't need a workspace API key.
To load the delivery API's HTTP spec into a tool, you can also use OpenAPI and the API catalog.
Example request
Add a genko.me blog list to this project.Check client configuration and list fetching in the genko.me docs.Use the generated lib/client.ts to connect, and handle the API key only on the server.Tell an empty list apart from a fetch error.Share your real API names and field layout as well, so the AI doesn't build on fields that don't exist. After you change something in the dashboard, update the generated types with genko pull.
Check after connecting
Being able to read the docs is not the same as your app being able to reach the delivery API. Your app needs a workspace ID (slug), an API key, and a published entry. You can confirm this in the Quickstart.
Even for AI-generated code, check that it doesn't put the API key on a public page, that it uses fields that exist, and that it handles 404s and empty values.
Last updated: October 6, 2026