# Port OS

> Discover connected devices and control their declared capabilities with scoped OAuth access.

## Start

- [For agents](/docs/for-ai/index.md): ordered setup and documentation tree.
- [Authentication and DCR](/docs/for-ai/guides/authentication-and-dcr.md): discovery, public registration, PKCE, callback, refresh.
- [Quickstart](/docs/for-ai/guides/quickstart.md): inspect one light and verify its reported state.

## Reference

- [Task guides](/docs/for-ai/guides/index.md)
- [OpenAPI](/docs/references/openapi.yaml)
- [Capabilities](/docs/for-ai/capability-interfaces.md)
- [Authorization](/docs/for-ai/guides/authorization-rules.md)
- [MCP reference](/docs/reference/mcp/): every MCP tool, what it does and the scope it needs.
- [Resources](/docs/resources/): public repositories, packages, skills and endpoints for building against Port OS.

## Blog

- [All posts](/blog/): articles on connecting AI assistants to devices; the guides above stay canonical.
- [Agentic API development: top 10 best practices](/blog/agentic-api-development-best-practices/): If you want AI to use your API, it has to be easy for them. In this list we go through 10 best practices for making agents work well across authentication, discoverability, documentation, and more.
- [How the Port OS MCP integration optimizes UX for Agents](/blog/how-the-port-os-mcp-integration-works/): Exposing agents directly to Port OS APIs can work, but leads to poor accuracy and confusion. We detail the skills, API changes, and custom integrations that we build for agents to work well with Port OS.