MCP Tools Reference
Orchestre provides eight specialized MCP tools that power its intelligent development capabilities. Each tool serves a specific purpose in the development workflow.
Available Tools
initialize_project
Smart project initialization with template selection and setup
View Documentation →Tool Categories
Project Setup
- initialize_project: Creates new projects from templates or adds Orchestre to existing projects
Analysis & Planning
- analyze_project: Understands requirements and complexity
- generate_plan: Creates actionable development plans
- research: Conducts AI-powered research on technical topics
Code Quality
- multi_llm_review: Performs comprehensive code review with consensus
Visual Tools
- take_screenshot: Capture full screen or specific windows
- get_last_screenshot: Retrieve recent screenshots
- list_windows: List available windows (macOS only)
How Tools Work
Invocation
Tools are called by Claude Code through the MCP protocol:
// Claude invokes a tool
mcp__orchestre__analyze_project({
requirements: "Build a real-time chat application",
context: {
template: "makerkit",
constraints: ["Must scale to 10k users"]
}
})Response Format
All tools return structured JSON responses:
{
"content": [
{
"type": "text",
"text": "{\"success\": true, \"data\": {...}}"
}
],
"isError": false
}Tool Composition
Tools work together in intelligent workflows:
Example workflow:
- Analyze requirements to understand complexity
- Generate a plan based on analysis
- Execute the plan with Claude's help
Error Handling
All tools implement consistent error handling:
{
"error": "Template 'invalid' not found",
"details": "Available templates: makerkit-nextjs, cloudflare-hono, react-native-expo",
"suggestion": "Use one of the available templates or provide a custom repository URL"
}Performance Considerations
| Tool | Typical Duration | Factors |
|---|---|---|
| initialize_project | < 1s | File I/O, template size |
| analyze_project | 2-5s | AI processing, complexity |
| generate_plan | 3-7s | Requirements complexity |
| multi_llm_review | 5-10s | Multiple AI calls, consensus |
Best Practices
1. Use Tools Appropriately
✅ Do:
- Use
analyze_projectbeforegenerate_plan - Call
initialize_projectto set up new or existing projects - Provide rich context for better results
❌ Don't:
- Skip analysis for complex projects
- Manually create .orchestre structure
- Provide minimal context
2. Handle Tool Responses
Always check for errors:
const result = await analyze_project(requirements);
if (result.isError) {
console.error(result.content[0].text);
// Handle error appropriately
}3. Provide Context
More context = better results:
// Minimal context
{ requirements: "Add auth" }
// Rich context
{
requirements: "Add JWT-based authentication",
context: {
template: "makerkit-nextjs",
constraints: ["Must support SSO", "Enterprise-grade security"]
}
}API Quotas
Tools that use external APIs have quotas:
| Tool | API | Free Tier | Rate Limits |
|---|---|---|---|
| initialize_project | None | N/A | None |
| analyze_project | Gemini | 60 QPM | 1M tokens/day |
| generate_plan | Gemini | 60 QPM | 1M tokens/day |
| multi_llm_review | Multi | Varies | Provider-specific |
Troubleshooting
Common Issues
Tool Not Available
- Check MCP server is running
- Verify tool registration in server.ts
- Restart Claude Code
API Errors
- Verify GEMINI_API_KEY is set
- Check API quotas
- Review error messages
Timeout Issues
- Complex operations may take time
- Consider breaking into smaller tasks
- Check network connectivity
Summary
Orchestre's MCP tools provide the foundation for intelligent development workflows. By understanding each tool's purpose and capabilities, you can build sophisticated applications with AI assistance.