pidster/persona-mcp

Personas MCP Server

A Model Context Protocol (MCP) server that enhances AI assistants with specialized personas for improved problem-solving in software development tasks.

Why Use Personas?

• Focused Solutions: Get targeted advice from the right perspective (architecture vs implementation vs debugging)
• Deep Expertise: Each persona brings specialized knowledge and problem-solving approaches
• Consistent Approach: Personas maintain consistent methodologies across conversations
• Faster Problem Solving: Skip generic advice and get straight to expert-level guidance
• Team Alignment: Use the same personas your team would consult in real life

Quick Start (5 minutes)

# 1. Clone and build
git clone https://github.com/pidster/persona-mcp.git
cd persona-mcp
npm install && npm run build

# 2. Add to Claude Desktop config (~/.claude/claude_desktop_config.json)
{
  "mcpServers": {
    "personas": {
      "command": "node",
      "args": ["/absolute/path/to/persona-mcp/dist/index.js"]
    }
  }
}

# 3. Restart Claude Desktop

# 4. Test in Claude: "Please adopt the architect persona"

For detailed setup instructions, see the .

Available Personas

Documentation

Getting Started

• - Detailed installation and setup
• - Get running in 5 minutes
• - Connect with Claude Desktop
• - How to effectively use personas

Customization

• - Build your own specialized personas
• - Integrate with JavaScript, Python, Go, Ruby

Technical Reference

• - Complete API documentation
• - Technical architecture overview
• - How persona matching works
• - OpenTelemetry metrics guide

Development

• - Set up your dev environment
• - How to contribute to the project
• - Code examples in multiple languages

Support

• - Frequently asked questions
• - Common issues and solutions
• - Future plans and features

Features

• Dynamic Persona Loading: Automatically loads personas from multiple sources
• Intelligent Recommendations: AI-powered persona matching for your tasks
• Multi-Factor Scoring: Advanced algorithm considering keywords, expertise, and complexity
• Full MCP Support: Resources, prompts, tools, and streaming responses
• REST API: Direct HTTP endpoints for non-MCP clients
• OpenTelemetry Metrics: Built-in performance monitoring

Examples

Using Personas in Claude

// Get recommendations
Which persona would be best for debugging a memory leak?

// Adopt a persona
Please adopt the debugger persona to help me troubleshoot this issue.

// Compare personas
Compare the architect and developer personas for API design.

REST API Usage

# Get all personas
curl http://localhost:3000/api/personas

# Get recommendations
curl -X POST http://localhost:3000/api/recommend \
  -H "Content-Type: application/json" \
  -d '{"query": "debug memory leak", "limit": 3}'

See more for different programming languages.

Configuring AI Assistants for Automatic Persona Selection

To enable your AI assistant to automatically select and adopt the most appropriate persona for each task, add instructions to your assistant's configuration file:

For Claude (CLAUDE.md)

When the Personas MCP server is available, automatically select and adopt the most appropriate persona for each task by:
1. Analyzing the user's request to identify the type of task (debugging, architecture, implementation, etc.)
2. Using the @recommend-persona tool to get persona recommendations
3. Adopting the highest-scoring persona for the task
4. Informing the user which persona was selected and why

Example: If a user asks about debugging a memory leak, automatically adopt the debugger persona.

For Other AI Assistants

Similar instructions can be added to:

• CURSOR.md for Cursor
• .github/copilot-instructions.md for GitHub Copilot
• Project-specific AI configuration files

See our for detailed examples.

Contributing

We welcome contributions! Please see our for details on:

• Setting up your development environment
• Adding new personas
• Submitting pull requests
• Code style guidelines

Getting Help

• Report Issues - Bug reports and feature requests
• - Common questions answered
• Contact: @pidster

License

MIT - See file for details