Crucis Documentation¶
~2 min read
Generate code that actually works. Crucis provides structured automated feedback — tests, adversarial review, constraints, holdout verification — so AI agents iterate until they get it right.
Get started¶
- Start Here — Install Crucis and run your first pipeline in under 3 minutes
- New Project Quickstart — Build a verified function from scratch — 5 min walkthrough
- Existing Codebase Quickstart — Add Crucis to your existing project with one command
- How It Works — Understand the fit → evaluate → verify pipeline
Learn more¶
- Why Crucis — See what makes Crucis different from testing alone
- Use Cases — From greenfield functions to legacy refactors — where Crucis fits
- MCP Server — Connect Crucis to Claude Code, OpenCode, or Codex
- CLI Commands — Every command, flag, and option in one place
Reference¶
- Objective Format — YAML schema for objectives
- Constraints — 34 AST-based code quality checks
- Configuration — environment variables and workspace settings
- Architecture — system design and module roles
- Background Optimizer — (experimental) automatic prompt improvement across runs
Help¶
- Troubleshooting — common errors and fixes
- FAQ — frequently asked questions
- Glossary — term definitions