Talk to your code. Ship faster.
code/chef is an AI-powered development team that lives in VS Code. Just type @chef in Copilot Chat and describe what you want—feature implementation, code reviews, infrastructure setup, CI/CD pipelines, or documentation. The AI team handles the rest.
@chef Add user authentication with JWT tokens and password reset
The Feature Dev agent writes production-ready code with tests.
@chef Review this PR for security vulnerabilities
The Code Review agent analyzes for security issues, performance, and best practices.
@chef Create a Docker Compose setup for my Node.js app with PostgreSQL
The Infrastructure agent generates Dockerfiles, compose files, and Terraform configs.
@chef Create a GitHub Actions workflow for testing and deployment
The CI/CD agent builds your pipelines across GitHub Actions, GitLab CI, Jenkins, and more.
@chef Document the API endpoints in this codebase
The Documentation agent creates README files, API docs, and architecture diagrams.
@chef Train a fine-tuned model for the Feature Dev agent
The Infrastructure agent handles the complete model lifecycle:
- Training: Fine-tune models on your codebase via HuggingFace AutoTrain
- Evaluation: Compare candidates vs baseline with 5 metrics (accuracy, completeness, efficiency, latency, integration)
- Deployment: Update agent models safely with automatic rollback
- A/B Testing: Measure improvement with comprehensive experiment tracking
VS Code Commands:
codechef.modelops.train— Start training wizard with cost estimationcodechef.modelops.evaluate— Evaluate model performance with LangSmithcodechef.modelops.deploy— Deploy model to agent (30s process)codechef.modelops.rollback— Rollback to previous version (<60s)codechef.modelops.modelVersions— View deployment history
Training Modes:
| Mode | Cost | Duration | Use Case |
|---|---|---|---|
| Demo | $0.50 | 5 min | Quick validation |
| Production | $3.50 | 60 min | Full training |
See LLM Operations Guide for complete documentation.
| Traditional Workflow | With code/chef |
|---|---|
| Switch between 5+ tools | One chat interface |
| Copy-paste context everywhere | AI understands your codebase |
| Wait for code reviews | Instant AI analysis |
| Manual documentation | Auto-generated docs |
| Complex CI/CD setup | Plain English commands |
code/chef automatically picks the right AI model for each task via OpenRouter:
| Task | Model | Cost/1M | Why |
|---|---|---|---|
| Orchestration | Claude 3.5 Sonnet | $3.00 | Best reasoning for routing |
| Code Generation | Qwen 2.5 Coder 32B | $0.07 | Purpose-built for coding |
| Code Review | DeepSeek V3 | $0.75 | Excellent analytical reasoning |
| Infrastructure | Gemini 2.0 Flash | $0.25 | Fast, 1M context for IaC |
| CI/CD | Gemini 2.0 Flash | $0.25 | Great at YAML generation |
| Documentation | DeepSeek V3 | $0.75 | Strong technical writing |
flowchart TB
subgraph VSCode["🖥️ VS Code"]
Chat["@chef Add JWT auth to my Express API"]
end
subgraph Orchestrator["🧑🍳 code/chef Orchestrator"]
Supervisor["Supervisor\n(Head Chef)"]
subgraph Agents["Specialized Agents"]
FeatureDev["🚀 Feature Dev"]
CodeReview["🔍 Code Review"]
CICD["⚡ CI/CD"]
Infra["🏗️ Infrastructure"]
Docs["📚 Documentation"]
end
Tools["🔧 150+ MCP Tools"]
end
subgraph Integrations["External Services"]
GitHub["🐙 GitHub"]
Linear["📋 Linear"]
Docker["🐳 Docker"]
Metrics["📊 Metrics"]
end
Chat --> Supervisor
Supervisor --> FeatureDev
Supervisor --> CodeReview
Supervisor --> CICD
Supervisor --> Infra
Supervisor --> Docs
FeatureDev --> Tools
CodeReview --> Tools
CICD --> Tools
Infra --> Tools
Docs --> Tools
Tools --> GitHub
Tools --> Linear
Tools --> Docker
Tools --> Metrics
- Download from GitHub Releases
- Install:
Ctrl+Shift+P→ "Extensions: Install from VSIX..." - Configure:
Ctrl+Shift+P→ "code/chef: Configure" - Use: Open Copilot Chat and type
@chef <your request>
That's it! See QUICKSTART.md for self-hosting options.
You: @chef Build a REST API for managing blog posts with CRUD operations
Chef: I'll create a complete blog API with:
- Express routes for posts (GET, POST, PUT, DELETE)
- PostgreSQL database schema
- Input validation
- Error handling
- Unit tests
[Creates files, runs tests, opens PR]
You: @chef Check my authentication code for security issues
Chef: I found 3 issues:
🔴 Critical: Password stored in plain text (line 45)
🟡 Warning: Missing rate limiting on login endpoint
🟢 Suggestion: Consider adding CSRF protection
[Links to specific lines with fix suggestions]
You: @chef Set up Kubernetes deployment for my microservices
Chef: I'll create:
- Deployment manifests for each service
- ConfigMaps and Secrets
- Horizontal Pod Autoscaler
- Ingress configuration
[Generates YAML files with best practices]
| Command | What it does |
|---|---|
@chef <task> |
Execute any development task |
@chef /status |
Check current task progress |
@chef /workflow <name> |
Run a specific workflow |
@chef /tools |
See available integrations |
code/chef connects to your existing tools:
- GitHub — PRs, issues, actions
- Linear — Project management, approvals
- Docker — Container management
- Databases — PostgreSQL, Redis
- Cloud — DigitalOcean, AWS (coming)
Want full control? Run your own code/chef instance.
git clone https://github.com/Appsmithery/code-chef.git
cd code-chef && cp config/env/.env.template config/env/.env
# Add your API keys to .env
cd deploy && docker-compose up -dSee QUICKSTART.md for detailed setup.
| Guide | Description |
|---|---|
| Quick Start | Installation & first steps |
| Architecture | How it works under the hood |
| Deployment | Production setup |
- Fork the repository
- Create feature branch:
git checkout -b feature/amazing-feature - Commit changes:
git commit -m 'Add amazing feature' - Push:
git push origin feature/amazing-feature - Open Pull Request
MIT License — see LICENSE
Built with ❤️ using LangGraph, MCP, and VS Code