L1 AI Access Protocol (MCP Binding)v0.1

The L1 AI Access Protocol defines how AI platforms connect to Open AIoT: an MCP (Model Context Protocol) binding that maps L0 capabilities into tools the AI can call directly — integrate once, and every compliant AI platform works.

Why MCP

MCP has become the de-facto standard for AI clients to consume external tools: Kimi, Claude Desktop, Cursor and other mainstream platforms all support it. Open AIoT does not invent yet another protocol — L1 is simply “the Open AIoT capability model bound onto MCP”, so AI platforms can connect with zero modification.

What the binding does

• Connection setup: the MCP server bridges the device side and the AI client, handling authentication and sessions.
• Capability mapping: L0 properties, actions and events are mapped to MCP tools and resources, which the AI discovers and calls in the standard way.
• Invocation and receipts: tool calls carry L2 authorization credentials; results and confirmation requests are returned in the specified format.

Highlights of v0.1

• The minimal contract of the MCP server binding: tool naming conventions and the serialization format of capability descriptors.
• Call flows for the three foundational tools: listing devices, reading state and executing actions.
• A unified return structure for errors and confirmation requests (working with L2 human-in-the-loop).

To get hands-on, read the 5-Minute Quickstart; MCP server downloads and deployment resources are on the developer page.