ai-agent-team

Overview

The AI Agent Team MCP server implements a Thread Manager system that provides persistent memory and semantic search capabilities for AI assistants. It features a layered memory architecture with 30-50% improved context accuracy through vector retrieval instead of time-based search. The server supports multiple specialized AI roles (product manager, frontend, backend, etc.) and enables multi-threaded task management with full context recovery. It includes local embedding models for offline operation and integrates with Git for automatic version control.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• pm-start

  Create a new product manager thread with a task description

  Try:Create a product manager thread for designing an e-commerce shopping cart featureTry:Start a product manager thread to analyze user authentication requirements
• fe-start

  Create a new frontend development thread with a task description

  Try:Create a frontend thread to implement a user dashboard UITry:Start a frontend development thread for the shopping cart interface
• be-start

  Create a new backend development thread with a task description

  Try:Create a backend thread for implementing user authentication APITry:Start a backend development thread for the shopping cart API
• qa-start

  Create a new QA testing thread with a task description

  Try:Create a QA thread for testing the payment processing functionalityTry:Start a QA testing thread for the new user registration flow
• threads

  List all available threads with their details

  Try:Show me all my active threadsTry:List all threads and their status
• thread-switchthread-switch <thread_id>

  Switch to a specific thread to resume its context

  Try:Switch to thread abc123 to resume the user authentication workTry:Change to thread xyz789 to continue the shopping cart development
• thread-info

  Show information about the current active thread

  Try:Show details of the current threadTry:Display information about this thread context
• changelog-generate

  Generate changelog from Git history

  Try:Generate a changelog for the entire projectTry:Create a release with GitHub Release integration
• softcopyright

  Generate software copyright application materials

  Try:Generate softcopyright materials for this React projectTry:Create software documentation for the Python application
• tidy-my-desktop

  Intelligently organize desktop files and directories

  Try:Help me organize my desktop filesTry:Clean up the download directory by file type
• drawnote

  Create visual notes and diagrams from text content

  Try:Create a colorful handwritten note about AI development historyTry:Generate a technical diagram using the tech innovation style

Comparable tools

Installation

# Step 1: Install globally
npm install -g ai-agent-team

# Step 2: Initialize configuration (global or project)
ai-agent-team init

# Step 3: Enable Thread Manager MCP server
claude mcp add thread-manager node "YOUR_PATH/.claude/skills/thread-manager/dist/index.js"

// Add to Claude Desktop config
{
  "mcpServers": {
    "thread-manager": {
      "command": "node",
      "args": ["/path/to/.claude/skills/thread-manager/dist/index.js"]
    }
  }
}

FAQ

Why is the /threads command not available?

The Thread Manager MCP server hasn't been enabled. Use 'claude mcp add thread-manager node <path>' to enable it and restart Claude Code.

What's the difference between global and project initialization?

Global initialization (~/.claude/) shares data across all projects while project initialization (./.claude/) keeps data isolated to the current project.

Compare ai-agent-team with