code-ai

Write a gated, multi-agent pipeline once — deploy it to any AI coding assistant, and run it through a bundled MCP orchestration server.

This monorepo holds code-ai: a CLI that installs agent/skill pipelines into the layout each AI tool expects, plus a Model Context Protocol (MCP) server that runs those pipelines gate by gate — each stage produces an artifact, gets classified, and waits for a human sign-off before the next stage runs.

👉 Full usage, install, and the MCP gate-pipeline docs live in the package README: packages/installer/README.md.

npm install -g code-ai-installer   # ships both the CLI and the MCP server
code-ai                            # interactive wizard

Package

Since DEV-100 (consolidation) the whole toolkit ships as a single npm package with two binaries:

Path	npm	Binaries
packages/installer	code-ai-installer (v4.x)	code-ai (CLI installer) · code-ai-mcp (MCP server)

Inside packages/installer/src/:

• index.ts, installer.ts, … — CLI installer logic (multi-target: Claude, Codex, Copilot, Qwen, Gemini/Antigravity, Kimi)
• mcp/ — MCP server runtime (gate state machine, decision log, Auditor tools, action tools)
• shared/ — zod schemas + TypeScript types consumed by both the CLI and the MCP server
• domains/<id>/ — the shipped pipelines: development, content, analytics, product (agents, skills, pipeline.yaml, base persona)

code-ai-mcp is registered automatically in .mcp.json when installing for target=claude. Legacy @denish1209/types@0.1.0 is deprecated on npm — superseded by code-ai-installer@4.0.0.

How it works (in one breath)

Each domain is a gated pipeline (development 8 gates, content 6, analytics 8, product 10). The conductor routes the run; at every gate the owning role fetches its skills (get_skill), produces a deliverable (submit_artifact), and a human signs off (sign_off) before advance_gate moves on — with a circuit breaker on repeated rollbacks. A meta Auditor reads aggregate run metrics and proposes improvements to the toolkit itself. See the package README for the full tool surface.

Workspace commands

npm install        # install workspace deps
npm run build      # build
npm test           # run all tests
npm run lint       # lint

License

MIT