Skip to content

LazyAI

Scaffold a canonical, multi-tool AI development environment from one CLI, with optional orchestration scaffolding and MCP runtime integration.

lazyai-cli initializes and maintains a tool-agnostic canonical layer under .ai/, then compiles it into native formats for OpenCode, Claude Code, and GitHub Copilot. It ships with bundled agents, skills, templates, rules, and optional MCP servers so teams can adopt a consistent AI operating system with minimal configuration.

What it does

  • One-time setup: lazyai-cli init scaffolds canonical files, tool-native directories, and an MCP catalog.
  • Compile: lazyai-cli compile regenerates per-tool configs from the canonical source of truth.
  • Update: lazyai-cli update refreshes library content while preserving customizations.
  • Doctor: lazyai-cli doctor checks drift, missing files, and skill state.
  • Orchestration (opt-in): lazyai-cli init --enable-servers orchestrator scaffolds chain/team/workflow definitions and registers the local orchestrator MCP runtime.

Where to start

  • Quick Start — run your first setup in minutes
  • Installation — install options and prerequisites
  • How It Works — canonical source, compile model, and manifest tracking
  • CLI Reference — full command and flag documentation
  • GitHub Wiki — short-form operational notes and release/install references

Architecture at a glance

lazyai-cli (Go binary)
   │ init ──▶ .ai/ (canonical)
   │ compile ──▶ .opencode/ + .claude/ + .github/ + .vscode/
   │ doctor ──▶ .ai-setup.json / .ai-setup.db (manifest + SQLite)
   │
   └── optional orchestrator MCP server
         └── lazyai-orchestrator (Go runtime)
             └── catalog, state, handoffs, prompt composition

The execution path uses local native agents (Claude Code, OpenCode, Copilot) directly. A2A is an optional config seam only; remote/network execution is not the default.

Status

  • CLI: Go module at github.com/rluisb/lazyai/packages/cli
  • Platforms: macOS, Linux
  • License: MIT