A production-ready multi-agent system for Claude Code that orchestrates specialized agents for different development tasks. Optimized for cost, quality, and speed.
| Agent | Model | Purpose |
|---|---|---|
| architect | opus | System design, execution plans, architecture decisions |
| product-manager | opus | Product strategy, prioritization, growth analysis |
| frontend | sonnet | React/Next.js UI development |
| python-backend | sonnet | FastAPI/Python backend development |
| deployment | sonnet | Railway deployments, cost management |
| qa | haiku | CI/CD validation, pre-commit checks |
| explore-light | haiku | Fast codebase exploration |
| ops | haiku | Server management, QA automation, git operations |
| Skill | Purpose |
|---|---|
| /explore | User-invocable fast codebase exploration |
-
Copy to your project:
cp -r .claude/ /path/to/your/project/ cp CLAUDE.md /path/to/your/project/
-
Customize for your stack (optional):
- Update
ops.mdwith your project paths - Update
qa.mdwith your CI commands - Update
deployment.mdfor your deployment platform
- Update
-
Start using agents:
# In Claude Code, the agents will be automatically available # Use Task tool with subagent_type to invoke them
The system uses a tiered model approach to minimize costs:
┌─────────────────────────────────────────────────────────────┐
│ MODEL COST TIERS │
├─────────────────────────────────────────────────────────────┤
│ │
│ HAIKU (1x baseline) - Use for: │
│ • explore-light: File search, code exploration │
│ • ops: Server management, QA automation, git │
│ • qa: CI/CD validation │
│ │
│ SONNET (~10x) - Use for: │
│ • frontend: UI implementation │
│ • python-backend: API implementation │
│ • deployment: Infrastructure management │
│ │
│ OPUS (~60x) - Use for: │
│ • architect: Strategic decisions │
│ • product-manager: Product strategy │
│ │
└─────────────────────────────────────────────────────────────┘
Expected savings: 30-40% vs. using expensive models for all tasks
User Request
│
├──► Product Manager (evaluate UX, define specs)
│
├──► Architect (create execution plan)
│
└──► [After plan approved]
│
├──► Frontend Agent ─────┐
│ ├──► QA Agent (test everything)
└──► Backend Agent ──────┘
│
▼
Deployment Agent (deploy)
- Stage-based design (Early/Growth/Scale)
- Creates execution plans for other agents
- Cost optimization framework
- Architecture Decision Records (ADRs)
Synthesizes methodologies from:
- Shreyas Doshi (LNO Framework, CPSR)
- Marty Cagan (Four Risks Framework)
- Sean Ellis (ICE Prioritization, PMF Test)
- Gibson Biddle (DHM Model)
- Julie Zhuo (Product Sense)
- Brian Balfour (Four Fits Framework)
- Next.js route conflict detection
- TypeScript validation
- Full CI simulation
- API contract validation
- Railway MCP integration
- Sleep/wake lifecycle for cost savings
- PR preview environments
- Production deployment with approval
The agents are configured for Next.js + FastAPI. To adapt:
- frontend.md - Change framework references
- python-backend.md - Switch to Django, Flask, Node.js, etc.
- ops.md - Update service commands
- qa.md - Update CI simulation
- deployment.md - Change from Railway to Vercel, AWS, etc.
Create .claude/agents/your-agent.md:
---
name: your-agent
description: What this agent does
tools: Read, Edit, Write, Bash, Grep, Glob
model: haiku | sonnet | opus
---
# Your Agent Name
[Instructions for the agent].claude/
├── agents/
│ ├── architect.md
│ ├── deployment.md
│ ├── explore-light.md
│ ├── frontend.md
│ ├── ops.md
│ ├── product-manager.md
│ ├── python-backend.md
│ └── qa.md
└── skills/
└── explore.md
CLAUDE.md # Project standards and workflow documentation
README.md # This file
MIT
Contributions welcome! Please submit issues and pull requests.