Open-source AI governance framework

AI governance that ships. Turn any repository into a governed AI workspace: policies, skills, agents, runbooks, and specs as versioned files. No hosted control plane, no vendor lock-in. One canonical chain across Claude Code, GitHub Copilot, OpenAI Codex, Gemini CLI, and Antigravity.

Prerequisites: Python 3.11+ and Git.

# pipx (recommended)
pipx install ai-engineering

# uv
uv tool install ai-engineering

# pip (inside a venv)
python -m venv .venv && source .venv/bin/activate
pip install ai-engineering

Verify:

ai-eng version

Update later with pipx upgrade ai-engineering (or uv tool upgrade / pip install --upgrade), then ai-eng update in each project followed by ai-eng doctor to verify.

cd your-project
ai-eng install .
ai-eng doctor

install scaffolds the governance root, detects your stack, mirrors skills to every configured IDE, and wires the secrets-gate. doctor validates the result.

Run ai-eng install --help for the full options matrix; the doctor output names every step that has to pass.

Telemetry is strict-opt-in (default disabled). The audit chain is local NDJSON; external emitters require explicit operator opt-in via .ai-engineering/manifest.yml > telemetry.*.

ai-engineering ships without a built-in memory layer. Engram is a peer product maintained by Gentleman-Programming/engram; install it separately if you want cross-session memory (spec-132 D-132-06; the installer no longer wires Engram automatically). See the Optional: Engram section of CLAUDE.md for OS-specific install commands and the IDE setup matrix.

This framework defines a single canonical chain that every supported IDE follows identically. The full ruleset, principles, surface index, and chain definition live in:

• AGENTS.md — canonical "how AI works in this repo" payload (mirrored byte-equivalent into CLAUDE.md, GEMINI.md, and .github/copilot-instructions.md)
• CONSTITUTION.md — project identity (mission, stakeholders, vocabulary, prohibitions, compliance gates)
• CHANGELOG.md — version history and breaking-change reference

The skill catalogue, agent roster, runbook list, quality-gate thresholds, and CLI command reference all live in AGENTS.md (canonical) and .ai-engineering/reference/cli-reference.md. They are not duplicated here.

ai-engineering builds on ideas, patterns, and principles from these projects:

Contributions are welcome. See CONTRIBUTING.md for development setup, code style, testing, and the pull request process.

This project follows the Contributor Covenant Code of Conduct. See CODE_OF_CONDUCT.md.

MIT. See LICENSE.