One source for every AI coding assistant

You edit one source; AGENTS.md, CLAUDE.md, .cursor/rules/, .github/instructions/, skill folders, MCP and settings files follow. AI-Rulez also validates that content, pins and signs it, and packages it as plugins.

How it works

You write rules, context, skills, agents and commands as Markdown files and list the tools you use in .ai-rulez/config.toml:

version = "5.0"
name = "acme"
presets = ["claude", "cursor", "codex", "copilot", "gemini"]

ai-rulez generate renders each tool’s own layout from that one source:

acme/
├── .ai-rulez/              # the source you edit
│   ├── config.toml
│   ├── rules/
│   ├── context/
│   └── skills/
├── AGENTS.md               # read by Codex, Cursor, Copilot and Gemini CLI
├── CLAUDE.md               # imports @AGENTS.md
├── .agents/skills/         # skills for tools that read .agents/skills
├── .claude/skills/         # skills for Claude Code
├── .gemini/settings.json   # points Gemini CLI at AGENTS.md
├── .mcp.json, .cursor/mcp.json, .vscode/mcp.json, .codex/config.toml
└── ...

Change the source, run generate again, and every tool stays in sync. ai-rulez generate --check fails in CI when the committed files drift from the source.

Get started

1. Install the CLI, or run it with npx ai-rulez@latest or uvx ai-rulez.
2. Follow the Quick start: init, choose presets, add a rule, generate.
3. Commit .ai-rulez/ together with the generated files.

Already have a CLAUDE.md, AGENTS.md or .cursor/rules/? ai-rulez init --from auto imports them. Upgrading a 4.x project? See Migrating to v5.

What you can do

Author

Write rules, context, skills, agents, commands and review checks; organize them with domains, profiles, roles and shared includes. Start with Configuration, Rules, Skill frontmatter and Domains and profiles.

Generate

Render native files for 52 assistants, per project or in your home directory, with hooks, permissions and MCP servers translated for each. See Supported harnesses, AGENTS.md and Hooks and settings.

Bundle and publish

Package plugins and marketplaces, Agent Plugins and ARD manifests, and release them to GitHub, npm or OCI. See Authoring plugins and Publish.

Validate

Lint configured projects, or plain AGENTS.md, SKILL.md and .claude/ files with no configuration, using stable AR rule codes; catch harness traps; run repository verifiers in CI and git hooks. See Strict validation, Harness traps, Verifiers and GitHub Actions.

Govern

Pin sources and content in a lock file, record reviewer approvals, sign with Sigstore, enforce an organization policy and produce an SBOM. See Trust model, Lock file and Signing.

Operate

Serve configuration and skills over MCP, search skills, run evals, improve skills, review changes with a model and export telemetry. See MCP server, Evals, Review and Judge.

The formats ai-rulez writes (AGENTS.md, Agent Skills, Agent Plugins, OKF, llms.txt, CycloneDX, SPDX, in-toto and others) are checked against each standard’s own schema or rules; Standards conformance lists the versions and known gaps.

Reference

CLI guideCommands by task, exit codes and JSON output.

CLI command referenceEvery command and flag, generated from the binary.

SchemaJSON schemas of the configuration and every --format json report.

ArchitectureHow a source becomes generated files, and the package layout.

Standards conformanceThe open formats ai-rulez writes and how each is checked.

Embedding (Go API)Load, validate, plan and apply generation from Go.

Getting help

• ai-rulez --help and ai-rulez <command> --help describe every command and flag.
• ai-rulez validate --explain AR001 explains a rule code and how to suppress it.
• ai-rulez doctor diagnoses a project (drift, removed presets, missing tools) without changing anything.
• Report problems on GitHub.

This documentation covers ai-rulez v5 (version = "5.0" in config.toml). The changelog lists every release.