mcp_cdh

peerjakobsen/mcp_cdh

3.2

If you are the rightful owner of mcp_cdh and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

The Discovery Copilot MCP Server enhances product discovery workflows with AI-powered tools for interview guide generation and opportunity solution tree management, based on Teresa Torres' Continuous Discovery Habits methodology.

Tools
5
Resources
0
Prompts
0

Discovery Copilot MCP Server

The Model Context Protocol (MCP) server for customer discovery and interview management

Transform your product discovery workflow with AI-powered tools for interview guide generation and opportunity solution tree management. Built on Teresa Torres' Continuous Discovery Habits methodology for product teams practicing evidence-based customer development.

🎯 What's New - Phase 1 Complete!

✅ Major Features Launched:

  • - Complete tree lifecycle management with professional visualizations
  • - Story-based interview guides following Torres' methodology

🔧 25+ MCP Tools Available - Full feature set ready for Claude Desktop, ChatGPT, and Amazon Q Developer

✨ Core Features

🌳 Opportunity Solution Tree Builder

Status: ✅ Complete | Tools: 20+ specialized tools |

Build, manage, and visualize opportunity solution trees through simple text commands in your AI assistant.

Key Capabilities:

  • Complete Tree Lifecycle - Create, manage, update, and delete entire opportunity trees
  • Hierarchical Structure - Add opportunities, solutions, and experiments with proper relationships
  • Professional Visualizations - Generate diagrams in PNG, SVG, and PDF formats
  • Multiple Export Options - Markdown reports, visual diagrams, and structured data exports
  • Persistent Storage - All data stored locally with uvx-compatible database paths

Perfect For: Product managers building discovery trees, teams mapping customer problems to solutions, stakeholder presentations with visual diagrams.

🎯 Interview Guide Generator

Status: ✅ Complete | Tool: generate_interview_guide |

Generate structured interview guides based on Teresa Torres' Continuous Discovery Habits methodology.

Key Capabilities:

  • Story-Based Questions - Focus on past behavior rather than future speculation
  • Phase-Specific Templates - Tailored for problem discovery, solution discovery, outcome mapping, assumption testing
  • Customer Segment Customization - Specialized for new users, active users, power users, churned users, trial users
  • Focus Area Specialization - Deep-dive questions for onboarding, workflows, pain points, competitive analysis
  • Multiple Output Formats - Compact summaries, full JSON guides, or formatted Markdown

Perfect For: UX researchers preparing customer interviews, product teams conducting weekly discovery, anyone following continuous discovery practices.

🛠️ Available MCP Tools

Core Tool Categories

  • (5 tools) - Create, list, get, update, delete trees
  • (5 tools) - Add, update, remove, list, search opportunities
  • (6 tools) - Track solutions and experiments with status updates
  • (4 tools) - Generate diagrams and comprehensive reports
  • (1 tool) - Comprehensive guide creation with customization
  • (4 tools) - Server status and connectivity testing

Quick Tool Reference

Most Used Tools:

  • discovery_tree_create - Start new opportunity trees
  • discovery_opportunity_add - Add customer problems to trees
  • discovery_solution_add - Link solutions to opportunities
  • discovery_tree_visualize - Generate professional diagrams
  • generate_interview_guide - Create customer interview guides
  • health_check - Verify server connectivity

🚀 Installation & Setup

Prerequisites

  • uvx (recommended) - The modern Python package runner
  • Python 3.12+ - Required for the server
  • Claude Desktop / ChatGPT / Amazon Q Developer - Any MCP-compatible AI assistant

Installation Methods

Method 1: uvx (Recommended)

The easiest way to use Discovery Copilot is with uvx, which runs the server directly from GitHub without local installation:

# No installation needed! uvx handles everything automatically
Method 2: Local Development

For contributors or advanced users who want to modify the server:

# Clone the repository
git clone https://github.com/peerjakobsen/mcp_cdh.git
cd mcp_cdh

# Install in development mode
pip install -e .

# Run the server
python -m mcp_cdh

🔧 Configuration

Claude Desktop Setup

Add this configuration to your Claude Desktop settings file:

Location:

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json

Configuration:

{
  "mcpServers": {
    "discovery-copilot": {
      "command": "/opt/homebrew/bin/uvx",
      "args": [
        "--from",
        "git+https://github.com/peerjakobsen/mcp_cdh",
        "mcp-cdh"
      ],
      "env": {
        "MCP_CDH_DB_PATH": "~/.local/share/discovery-copilot/data.db"
      }
    }
  }
}

ChatGPT Desktop Setup

For ChatGPT Desktop with MCP support:

{
  "mcpServers": {
    "discovery-copilot": {
      "command": "/opt/homebrew/bin/uvx",
      "args": [
        "--from",
        "git+https://github.com/peerjakobsen/mcp_cdh",
        "mcp-cdh"
      ],
      "env": {
        "MCP_CDH_DB_PATH": "~/.local/share/discovery-copilot/data.db"
      }
    }
  }
}

Amazon Q Developer Setup

For Amazon Q Developer Chat:

{
  "mcpServers": {
    "discovery-copilot": {
      "command": "/opt/homebrew/bin/uvx",
      "args": [
        "--from",
        "git+https://github.com/peerjakobsen/mcp_cdh",
        "mcp-cdh"
      ],
      "env": {
        "MCP_CDH_DB_PATH": "~/.local/share/discovery-copilot/data.db"
      }
    }
  }
}

Environment Variables

Customize the server behavior with these environment variables:

  • MCP_CDH_DB_PATH - Custom database location (default: auto-detected)
  • MCP_CDH_LOG_LEVEL - Logging level: DEBUG, INFO, WARNING, ERROR (default: INFO)

💡 Quick Start Examples

Interview Guide Generation

Create an interview guide for problem discovery with active users focusing on onboarding pain points.

MCP Tool Call:

{
  "tool": "generate_interview_guide",
  "arguments": {
    "discovery_phase": "problem_discovery",
    "customer_segment": "active_users",
    "focus_areas": ["onboarding", "pain_points"],
    "interview_length": "30-45 minutes"
  }
}

Opportunity Solution Tree Building

Create a new opportunity tree for mobile checkout optimization, then add the main customer problem we discovered.

Step 1 - Create Tree:

{
  "tool": "discovery_tree_create",
  "arguments": {
    "name": "Mobile Checkout Optimization",
    "description": "Discovery tree for improving mobile checkout conversion",
    "outcome": "Increase mobile checkout completion by 25%"
  }
}

Step 2 - Add Opportunity:

{
  "tool": "discovery_opportunity_add",
  "arguments": {
    "tree_id": "tree_mobile_checkout_2025",
    "title": "Users abandon checkout on mobile devices",
    "description": "High abandonment rate observed in mobile analytics"
  }
}

Visual Documentation

Generate a professional diagram of our discovery tree for the stakeholder presentation.

MCP Tool Call:

{
  "tool": "discovery_tree_visualize",
  "arguments": {
    "tree_id": "tree_mobile_checkout_2025",
    "format": "png",
    "save_to_file": true,
    "filename": "checkout-optimization-tree.png",
    "theme": "professional"
  }
}

🎬 Real-World Scenarios

Ready to see Discovery Copilot in action? Check out these comprehensive workflow examples:

  • New Feature Discovery - 4-week sprint from research to recommendations
  • Customer Churn Investigation - 3-week analysis with evidence-based solutions
  • Competitive Response Strategy - 2-week competitive analysis and differentiation
  • Feature Sunset Decision - 6-week analysis and migration planning

Each scenario shows the complete tool sequence with actual MCP calls and expected outputs.

📚 Methodology & Best Practices

Discovery Copilot implements Teresa Torres' Continuous Discovery Habits methodology with tools designed for weekly customer interviews and evidence-based product decisions.

Core Principles

  • ✅ Story-Based Questions - "Tell me about the last time you..." (actual experiences)
  • ❌ Avoid Future Speculation - "Would you use..." (hypothetical opinions)
  • 🎯 Weekly Interview Cadence - Consistent customer contact for continuous learning
  • 🌳 Opportunity Solution Trees - Visual mapping of problems to solutions and experiments

Discovery Phases & Segments

4 Discovery Phases: Problem Discovery, Solution Discovery, Outcome Mapping, Assumption Testing 5 Customer Segments: New Users, Active Users, Power Users, Churned Users, Trial Users 6 Focus Areas: Onboarding, Feature Usage, Pain Points, Workflow, Decision Making, Competitive Analysis

Output Formats

  • Compact (default) - Lightweight summaries for AI assistants
  • JSON - Structured data for tool integration
  • Markdown - Human-readable guides with formatting
  • Visual Diagrams - Professional PNG/SVG/PDF exports

🔧 Troubleshooting

Common Issues

MCP Server Not Connecting:

  1. Verify uvx is installed: uvx --version
  2. Check configuration file syntax (JSON must be valid)
  3. Restart your AI assistant (Claude Desktop/ChatGPT/Amazon Q)
  4. Check the assistant's MCP logs for error messages

Assistant Crashes When Calling Tools:

  1. Try the ping tool first - if it works, the server is functioning
  2. Use server_info for basic information (lightweight response)
  3. Avoid debug_server_info unless specifically needed for debugging
  4. Restart your AI assistant if crashes persist

Database Permissions:

  1. Ensure write permissions for the database directory
  2. Try setting a custom MCP_CDH_DB_PATH in a writable location
  3. On macOS, check System Preferences → Security & Privacy

uvx Installation Issues:

# Install uv first
curl -LsSf https://astral.sh/uv/install.sh | sh

# Then uvx is available automatically
uvx --version

Path Issues:

  • Use full path to uvx: /opt/homebrew/bin/uvx (macOS) or /usr/local/bin/uvx (Linux)
  • Find uvx location: which uvx

Debug Mode

Enable debug logging by setting the environment variable:

{
  "env": {
    "MCP_CDH_LOG_LEVEL": "DEBUG",
    "MCP_CDH_DB_PATH": "~/.local/share/discovery-copilot/data.db"
  }
}

Testing the Server

You can test the server directly:

# Test health check
echo '{"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {"name": "health_check", "arguments": {}}}' | uvx --from git+https://github.com/peerjakobsen/mcp_cdh mcp-cdh

# Test ping
echo '{"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {"name": "ping", "arguments": {}}}' | uvx --from git+https://github.com/peerjakobsen/mcp_cdh mcp-cdh

🤝 Contributing

Discovery Copilot is built for the product community. Contributions are welcome!

Development Setup

# Clone and setup
git clone https://github.com/peerjakobsen/mcp_cdh.git
cd mcp_cdh

# Install dependencies
pip install -e .
pip install -e ".[dev]"

# Run tests
pytest

# Format code
black .
ruff check --fix

Adding New Templates

Interview templates are located in src/mcp_cdh/templates/. Follow the existing pattern for consistency with Torres' methodology.

📄 License

MIT License - see LICENSE file for details.

🙏 Acknowledgments

  • Teresa Torres for the Continuous Discovery Habits methodology
  • Model Context Protocol team for the MCP standard
  • FastMCP for the excellent MCP framework

📖 Documentation

Feature Guides

  • - Complete tree management, visualization, and export
  • - Story-based interview guide creation and customization
  • - Real-world workflows with step-by-step examples

API Reference

  • - Complete documentation for all 25+ MCP tools

Getting Started

  1. Install Discovery Copilot using uvx (recommended) or local development
  2. Configure your AI assistant (Claude Desktop, ChatGPT, or Amazon Q Developer)
  3. Try the quick examples to verify everything works
  4. for comprehensive workflows
  5. to understand Torres' principles

Ready to transform your product discovery workflow? Install Discovery Copilot and start building evidence-based opportunity solution trees and conducting story-based customer interviews! 🚀