Caliber
Caliber is a developer tool that automatically generates and maintains AI context files like CLAUDE.md, AGENTS.md, and Cursor rules, so they stay accurate as your code evolves. It solves the problem of stale, hand-written configs that cause AI agents to hallucinate paths and give outdated advice. It's for developers and teams using AI coding assistants like Claude Code, Cursor, Codex, OpenCode, or GitHub Copilot.
✨ Key features
- Generates configs for multiple AI platforms from one setup.
- Deterministic scoring audits config quality without LLM calls.
- Auto-refresh hooks keep docs in sync on every commit.
- Session learning captures patterns into CALIBER_LEARNINGS.md.
- MCP server discovery auto-configures for Claude and Cursor.
- Fully reversible with backups, undo, and dry-run previews.
🎯 Use cases
- Automatically generate and maintain CLAUDE.md for a project.
- Keep Cursor rules and AGENTS.md updated as code changes.
- Score and compare AI config quality across branches.
- Onboard new team members with consistent AI context.
- Learn from AI coding sessions to improve future guidance.
📦 Installation
🧰 Requirements: Requires Node.js >= 20. No API key needed for bootstrap and scoring; generation uses your existing Claude Code or Cursor subscription or your own Anthropic, OpenAI, MiniMax, or Vertex AI key.
npx @rely-ai/caliber bootstrap
🚀 Usage
npx @rely-ai/caliber bootstrap/setup-caliber❓ FAQ
Does Caliber overwrite my existing configs?
No. Caliber shows a diff of every proposed change and you accept, refine, or decline each one. Originals are backed up automatically.
Does Caliber need an API key?
Bootstrap and scoring run 100% locally with no LLM and no API key. Generation uses your existing Claude Code or Cursor subscription (no key) or you can bring your own key for Anthropic, OpenAI, MiniMax, or Vertex AI.
What's the difference between bootstrap and init?
Bootstrap installs agent skills in 2 seconds, then your agent runs /setup-caliber. Init is the full interactive CLI wizard. Both end up in the same place.
Does Caliber send my code anywhere?
Scoring is fully local. Generation sends a project summary (not source code) to your configured LLM provider. Anonymous usage analytics are collected but can be disabled with --no-traces or CALIBER_TELEMETRY_DISABLED=1.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.