AI Agents & MCP

RustaSea's predictable structure and conventions make it a good fit for AI coding agents. Every project can expose its knowledge surface over the Model Context Protocol:

cargo artisan mcp:serve

This is gated behind the ai-mcp (or mcp) cargo feature on the rustasea crate. A build without that feature does not register the command.

What the server exposes

The MCP server advertises tools and resources rather than raw file access.

Safety boundaries

Agentic development loop

sequenceDiagram
    participant Agent as Coding agent
    participant MCP as rustasea mcp:serve
    participant App as Booted application
    Agent->>MCP: route:list
    MCP->>App: inspect live route table
    App-->>MCP: Method, URI, Name, Action, Middleware, Binding
    MCP-->>Agent: route metadata
    Agent->>MCP: make:plan
    MCP-->>Agent: scaffold plan
    Agent->>App: generate controller, model, migration

Related generators

Agent-facing scaffolds live in the same CLI:

cargo artisan make:agent {name}
cargo artisan make:tool {name}