Quick Start
This guide walks you through initializing a project and running your first AI-assisted workflow.
Prerequisites
- Raven installed (Installation)
- At least one AI agent CLI installed:
claude,codex, orgemini - A git repository to work in
1. Initialize a Project
cd my-project
raven init
This creates a raven.toml configuration file. Open it and configure your project:
[project]
name = "my-project"
language = "go"
[agents.claude]
command = "claude"
model = "claude-sonnet-4-6"
effort = "high"
2. Implement Your First Task
Run the implementation loop for a phase:
raven implement --agent claude --phase 1
Use --dry-run to preview the generated prompt without invoking the agent:
raven implement --agent claude --phase 1 --dry-run
3. Run a Review
Run multi-agent review on your changes:
raven review --agents claude,codex --base main
4. Fix and Create a PR
Apply fixes from the review, then open a pull request:
raven fix --agent claude --report review-report.md
raven pr --base main --agent claude
Run the Full Pipeline
Or run everything in one command:
raven pipeline --phase 1 --impl-agent claude --review-agent codex --fix-agent claude
Launch the Dashboard
For an interactive view of running workflows:
raven dashboard
Next Steps
- Command Reference -- full flag tables and usage examples
- Configuration Reference -- all
raven.tomloptions - Architecture -- how Raven is designed