camille.run
Getting Started with Multi-Agent Collaboration
Set up multiple AI agents working together on your projects in under 5 minutes.
Prerequisites

Before you begin, ensure you have:

✅ Camille.run account
✅ Claude Code installed
✅ At least one project created in Camille
✅ MCP server configured
Quick Start (5 Minutes)

Install the wshobson/agents package with 95+ specialized agent types:

# Via Claude Code marketplace
/install claude-code-workflows

# Or manually
git clone https://github.com/wshobson/agents ~/.claude/plugins/agents

Verify installation:

ls ~/.claude/plugins/marketplaces/claude-code-workflows/plugins/*/agents/

You should see directories for: backend-development, frontend-development, security, testing, etc.

Camille's MCP server should already be configured if you followed the initial setup. Verify:

// ~/.claude/mcp-servers.json
{
  "camille": {
    "url": "https://camille.run/mcp/camille",
    "auth": {
      "type": "oauth2",
      "authorization_url": "https://camille.run/oauth/authorize",
      "token_url": "https://camille.run/oauth/token"
    }
  }
}

In Camille.run, create a new issue for your feature:

Issue: "Implement user authentication system"

Description: "Add JWT-based authentication with login, registration, password reset, and refresh token support"

Project: your-project

Open multiple Claude Code instances and run the agents:

Terminal 1 (Backend)
claude-code --agent backend-architect

# In Claude Code:
@camille analyze issue #123
Terminal 2 (Frontend)
claude-code --agent frontend-architect

# In Claude Code:
@camille analyze issue #123
Terminal 3 (Security)
claude-code --agent security-engineer

# In Claude Code:
@camille analyze issue #123
Terminal 4 (Testing)
claude-code --agent test-automator

# In Claude Code:
@camille analyze issue #123

Each agent will:

  • 1. Join the agent session automatically
  • 2. Analyze the issue from their expertise
  • 3. Create specialized todos
  • 4. Start working on assigned tasks
  • 5. Sync in real-time via WebSocket
What you'll see:
🤖 backend-architect joined session
   ✓ Created: "Design authentication API endpoints"
   ✓ Created: "Implement JWT token generation"
   ✓ Created: "Create user registration endpoint"

🤖 frontend-architect joined session
   ✓ Created: "Build login form component"
   ✓ Created: "Add authentication state management"
   ✓ Created: "Create protected route wrapper"

🤖 security-engineer joined session
   ✓ Created: "Add password hashing with bcrypt"
   ✓ Created: "Implement CSRF protection"
   ✓ Created: "Add rate limiting to auth endpoints"

🤖 test-automator joined session
   ✓ Created: "Write authentication E2E tests"
   ✓ Created: "Add unit tests for JWT service"
   ✓ Created: "Create integration tests"

⚠️ Conflict detected: Duplicate todo
   ✓ Auto-resolved: Merged duplicates

🎉 Session completed: 16 todos created, 0 conflicts
Common Workflows
# 1. Create issue
Issue: "Add payment processing"

# 2. Start agents (4 terminals)
Terminal 1: backend-architect
Terminal 2: frontend-architect
Terminal 3: security-engineer
Terminal 4: test-automator

# 3. Analyze issue
All agents: @camille analyze issue #456

# 4. Agents create todos
- Payment API endpoints (backend)
- Payment form UI (frontend)
- PCI compliance review (security)
- Payment flow E2E tests (test)

# 5. Work in parallel
All agents work simultaneously

# 6. Review & merge
Code reviews, then merge to main
# 1. Create issue
Issue: "Fix checkout cart calculation bug"

# 2. Start agents (3 terminals)
Terminal 1: debugger
Terminal 2: backend-architect
Terminal 3: test-automator

# 3. Root cause analysis
debugger: @camille analyze issue #789
- Identifies: Tax calculation logic error

# 4. Fix implementation
backend-architect: Implements fix
test-automator: Adds regression tests

# 5. Verification
Run tests, verify fix
# 1. Create issue
Issue: "Review authentication architecture"

# 2. Start agents (4 terminals)
Terminal 1: system-architect
Terminal 2: security-auditor
Terminal 3: backend-architect
Terminal 4: performance-engineer

# 3. Analyze from different angles
system-architect: Overall design
security-auditor: Security vulnerabilities
backend-architect: Implementation quality
performance-engineer: Performance impact

# 4. Consolidate findings
Agents create todos for improvements

# 5. Prioritize & implement
Execute improvements in order
Most Common Agent Types
Backend Development
  • backend-architect
    - API design, system architecture
  • database-architect
    - Schema design, query optimization
  • graphql-architect
    - GraphQL APIs, resolvers
Frontend Development
  • frontend-architect
    - Component architecture, state management
  • frontend-developer
    - UI implementation
  • ui-ux-designer
    - Design systems, accessibility
Security & Compliance
  • security-engineer
    - General security implementation
  • security-auditor
    - Security reviews, penetration testing
Testing & Quality
  • test-automator
    - Test automation, E2E tests
  • tdd-orchestrator
    - Test-driven development
  • quality-engineer
    - Code quality, testing strategy
Best Practices
🎯 Do's
  • ✅ Use descriptive issue titles
  • ✅ Limit to 3-8 agents
  • ✅ Choose complementary agents
  • ✅ Review conflicts
  • ✅ Monitor activity feed
  • ✅ Use agent_description
  • ✅ Clean up sessions
❌ Don'ts
  • ❌ Don't run duplicate agent types
  • ❌ Don't ignore heartbeat failures
  • ❌ Don't create vague issues
  • ❌ Don't disable conflict detection
  • ❌ Don't leave sessions open indefinitely
  • ❌ Don't mix personal and team work
Troubleshooting

Symptoms: Agents create duplicate todos, conflicts not detected

Solution:

# 1. Check WebSocket connection
# In each Claude Code instance
echo.connector.pusher.connection.state
# Should show: "connected"

# 2. Verify channel subscription
echo.connector.pusher.channels.channels
# Should include: "private-session.42"

# 3. Restart Reverb if needed
php artisan reverb:restart

Symptoms: Constant conflict detection messages

Solution:

Use more specific todo titles:
❌ Bad: "Fix bug"
✅ Good: "Fix cart tax calculation for EU customers"
Use agent_description field to prevent overlap
Next Steps
⚡ Deep Dive: WebSocket Sync

Understand the technical details of real-time synchronization

📚 Full Documentation

Complete overview of multi-agent collaboration features

Ready to experience the future of AI collaboration?

Create your first multi-agent session and see the magic happen.

Go to Dashboard →