Five Laps
Every task races through objective, product, design, plan, and execution. No more jumping straight to code.
vibe-racerYour AI race engineer
Five laps from objective to shipped code — you call the pit stops.
Skip to content
What is vibe-racer?
Your project is the car. Every pipeline stage is a lap. Every human checkpoint is a pit stop. vibe-racer is the race engineer in your ear — it does the heavy work, calls the strategy, and hands you the final say on every decision.
LLMs generate code without product thinking. There's no requirements gathering, no design review, no structured planning. You get code, but not necessarily the right code.
vibe-racer fixes this by enforcing a structured development process where the race engineer handles the heavy lifting at each lap — generating specs, asking questions, writing plans — while you retain control at every pit stop.
Quick Start
bash
npm install -g vibe-racer
cd your-project
vibe-racer init
vibe-racer new "Add user authentication"
# Write your objective, tick the checkbox, then:
vibe-racer driveThree commands to your first AI-assisted development cycle. Full guide here.
10 Things You Need to Know
1. What it is
vibe-racer is an AI-powered development pipeline that races tasks through five laps: objective, product, design, plan, and execution. It uses Claude Code sessions with specialized personas at each lap to produce structured, high-quality software.
2. The problem
LLMs generate code without product thinking. There's no requirements gathering, no design review, no structured planning. You get code, but not necessarily the right code. The gap between "generate code" and "build software" is where bugs, scope creep, and rework live.
3. The solution
A 5-lap pipeline that forces proper software development process before a single line of code is written. The race engineer does the heavy lifting at each lap — generating specs, asking questions, writing plans — while you retain control at every pit stop.
4. Pit stops by design
You approve every lap transition via a simple checkbox. No black-box automation. The race engineer cannot advance to the next lap without your explicit sign-off. If it needs more information, it radios back with follow-up questions and waits.
5. Role-based AI personas
Each lap uses a specialized persona for higher quality output. A Senior Product Designer handles objective and product review. A Software Architect handles design. A Software Engineer handles planning and execution. The right expertise at the right lap.
6. Security-first
A multi-layered guard system constrains every AI session: path containment, bash command blocklist, interpreter-aware network detection, dotenv protection, pre-commit secret scanning, and an append-only audit log. Not a sandbox, but a strong guardrail with documented limitations.
7. Git-native
Everything is local files, git branches, and commits. No external dependencies, no cloud lock-in, no database. State lives in YAML files alongside the plan documents. Branches are created per task, commits happen after each lap. You control when to push and merge.
8. Pre-filled answers
The race engineer doesn't just ask questions — it pre-fills its recommended answers to every question. You review them and only edit what you disagree with. This dramatically reduces the time spent at pit stops while keeping you in full control.
9. Zero-config start
Three commands to your first AI-assisted development cycle:
bash
vibe-racer init
vibe-racer new "my feature"
vibe-racer driveNo configuration files to write, no integrations to set up. Just an Anthropic API key and a project directory.
10. How to get started
Install with npm install -g vibe-racer, set your ANTHROPIC_API_KEY, and run your first task in under 5 minutes. The included example task lets you experience the full pipeline immediately. Read the full guide.