Browse the docs ▾

Agent integration

Discovery and machine-readable docs

The files and endpoints a machine can read to learn about Opmiz, and a plain account of what these docs can't do on their own.

What you can read

Docs map

https://docs.opmiz.com/llms.txt is a map of these docs in llms.txt format: every page, with one line about it.

Full docs

https://docs.opmiz.com/llms-full.txt is the whole of these docs in one Markdown file.

Sitemap

https://docs.opmiz.com/sitemap.xml lists every page for search engines.

REST API

https://api.opmiz.com/v1/openapi.json describes the API in OpenAPI 3.1. It's public, with no key.

MCP sign-in

https://mcp.opmiz.com/.well-known/oauth-protected-resource gives the MCP server's sign-in metadata and permissions.

Sign-in endpoints

https://mcp.opmiz.com/.well-known/oauth-authorization-server lists them.

The MCP tool list

Signing in gives the tools with their schemas and hints, and the server's instructions. The list holds only what the connection's permissions allow.

Retrieving these docs

The llms-full.txt file suits a single fetch, but it's large. For one answer, fetch llms.txt and then just the page you need. Scripts using curl or Python can fetch only the two llms files, not /docs/ pages, so use llms-full.txt from those. Every page has a stable address under /docs/. Pages are written so that each stands alone and names its tools, which also helps semantic search.
The order of authority is: what a tool returns, then the tool's description, then these docs. If they disagree, trust the tool, and tell us at support@opmiz.com.

How annotations help

Each tool states whether it is read-only, whether it is destructive, whether repeating it is safe and whether it reaches outside Opmiz. Clients can use these to decide what to run without asking. Tools that change the live site are marked destructive, so many clients ask the person first.

What's distributed, and what isn't

Publishing documentation doesn't put Opmiz in every AI agent. An agent knows about Opmiz only through one of the routes below.

Works today

A person adds the server address to their client. An agent or search tool finds these docs on the web.

MCP Registry

Opmiz is listed as com.opmiz/opmiz in the official MCP Registry, which directories and clients read.

Claude Code plugin

Installs the connection and four skills. See Skills and instructions.

Your own instructions

You add the instructions and skill to your project.

Not listed yet

Claude's connector directory, Claude Code's plugin directory and the ChatGPT app directory. We add each here once it is live.

Model knowledge

A model's training data may not include Opmiz, so don't assume that it knows these tools until it has read them.

For tool builders

If you run an agent framework or directory, the address and metadata above are all you need to describe Opmiz. Clients can register themselves through dynamic client registration or a client ID metadata document, so there's no approval step on our side. See Sign-in and permissions.

Fetch the map

curl https://docs.opmiz.com/llms.txt

Read the sign-in metadata

curl https://mcp.opmiz.com/\
.well-known/oauth-protected-resource

Read the API description

curl https://api.opmiz.com/v1/openapi.json

Find it in the MCP Registry

curl "https://registry.modelcontextprotocol.io\
/v0.1/servers?search=com.opmiz"