⚠️ DEVELOPMENT STATUS: PRE-ALPHA ⚠️

This project is currently in early development and not ready for production use. Tools are being actively developed and may change significantly before the first release.

A comprehensive Model Context Protocol (MCP) server combining file system tools, AI Librarian, code comprehension, project scaffolding, and the Think Tool.

Overview

The AI Dev Toolkit enhances AI assistants like Claude with powerful capabilities:

File System Tools: Read, write, and navigate the file system
AI Librarian: Persistent code comprehension system that maintains project context
Project Starter: Project generation and scaffolding
Think Tool: Structured reasoning for complex problems
Context Compression: Store and retrieve conversation history (Coming Soon)

Quick Start

Prerequisites

Python 3.8 or higher
Claude Desktop (or other MCP-compatible AI assistant)
Administrator privileges (for Claude Desktop integration on Windows)

Installation

# Clone the repository
git clone https://github.com/[your-github-username]/ai-dev-toolkit.git
cd ai-dev-toolkit

# Create a virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Run the installation script (coming soon)
python install.py

Note: The installation script is still in development. Currently, you'll need to manually start the server using the provided scripts.

Connect to Claude Desktop (Windows)

We provide a simple script to automatically connect the toolkit to Claude Desktop:

Run connect_to_claude.bat (requires administrator privileges)
The script will:
- Find your Claude Desktop installation
- Update the configuration to include the MCP server
- Restart Claude Desktop if needed

For manual configuration:

Open Claude Desktop
Go to Settings > MCP Servers
Click "Add Server"
Enter the following details:
- Name: AI Dev Toolkit
- URL: http://localhost:8000
Click "Save"
Grant permissions when prompted

Start the Server

# Start the server
python src/server.py

Using the Tools

In Claude Desktop, you can access the tools by typing:

@AI Dev Toolkit

Available Tools

File System Tools

read_file: Read file contents
write_file: Write to a file
edit_file: Make line-based edits
create_directory: Create a directory
list_directory: List directory contents
directory_tree: View directory structure
move_file: Move or rename files
search_files: Find files matching a pattern
get_file_info: Get file metadata

AI Librarian Tools

initialize_librarian: Set up the AI Librarian
query_component: Find component details
find_implementation: Locate code implementations
generate_librarian: Generate the librarian files

Project Starter Tools

create_project_plan: Generate a project plan
generate_project_structure: Create directory structure
create_starter_files: Generate starter code
setup_github_repo: GitHub setup instructions

Think Tool

think: A scratchpad for reasoning through problems

Coming Soon

Context Compression: Store and retrieve conversation history in an AI-optimized format
RAG Integration: Connect to vector databases and knowledge bases for enhanced context

Examples

Using the AI Librarian

# First, initialize the AI Librarian for a project (creates persistent context)
initialize_librarian("D:/Projects/my-project")

# Now Claude will maintain awareness of your codebase across conversations
# You can query components directly
query_component("D:/Projects/my-project", "MyClass")

# Search for implementations
find_implementation("D:/Projects/my-project", "connect_database")

# The AI Librarian automatically monitors your codebase for changes
# No manual updates needed - Claude always stays in sync with your code

Creating a New Project

# Create a project plan
create_project_plan(
    project_name="TaskMaster",
    project_description="A task management application",
    project_type="web",
    key_features=["User authentication", "Task creation and management", "Categories and tags"]
)

# Generate the project structure
generate_project_structure(
    structure_text="taskmaster/\n├── src/\n│   ├── components/\n...",
    output_directory="D:/Projects/taskmaster"
)

# Create starter files
create_starter_files(
    project_directory="D:/Projects/taskmaster",
    project_name="TaskMaster",
    project_type="web"
)

Context Compression (Coming Soon)

The Context Compression tool will:

Store conversation history in a format optimized for AI language models
Save compressed context in .ai_reference as JSON files
Allow continuity between sessions without manual context copying
Prioritize project-relevant information

Benefits:

Dramatically improved context retention between sessions
Better understanding of project history
More consistent assistance

RAG Integration (Future)

Future versions will include:

Connection to vector databases
Document embedding and retrieval
Semantic search across project documentation
Integration with knowledge management systems

Development

Project Structure

ai-dev-toolkit/
├── .ai_reference/       # AI Librarian self-reference system
├── docs/                # Documentation
│   ├── project-plan.md   # Project plan and design document
│   ├── architecture/     # System architecture documentation
│   └── images/           # Documentation images
├── src/
│   ├── server.py         # Main MCP server implementation
│   ├── librarian/        # AI Librarian components
│   ├── mcp/              # MCP protocol components
│   └── utils/            # Utility functions
├── scripts/
│   ├── connect_to_claude.bat       # Claude Desktop connection script
│   ├── run_project_generator.bat   # Project generation runner
│   ├── project-generator-script.py  # Project generation implementation
│   ├── run_librarian_generator.bat # AI Librarian update script
│   └── context-compressor.py       # Context compression tool
├── tests/                # Test cases
├── requirements.txt      # Project dependencies
├── install.py           # Installation script
└── README.md             # This file

Running Tests

# Run tests
python -m pytest tests/

Development Roadmap & Timeline

Feature	Status	Expected Completion
File System Tools	In Progress	Q2 2025
AI Librarian	Planning	Q2 2025
Project Starter	Concept	Q3 2025
Think Tool	Concept	Q3 2025
Context Compression	Planned	Q4 2025
RAG Integration	Future	2026

How Tools Connect to Claude

This project uses the Model Context Protocol (MCP) to connect tools to Claude:

The MCP server runs locally on your machine
Claude Desktop connects to the server via localhost
Once connected, Claude can "see" and invoke the available tools
Tools execute on your local machine and return results to Claude
Claude interprets the results and communicates them back to you

MCP Architecture

Note: The architecture diagram will be added in a future update.

Troubleshooting

Claude Desktop Connection

If the automatic connection script fails:

Check if Claude Desktop is running and close it
Try running the script with administrator privileges
Manually edit the Claude config at: %APPDATA%\Claude\config.json
Ensure the server is running before connecting Claude

Server Issues

If the server fails to start:

Check if another process is using port 8000
Verify Python version is 3.8 or higher
Make sure all dependencies are installed
Check logs for specific error messages

Contributing

Contributions are welcome! Please feel free to submit pull requests.

License

This project is licensed under the MIT License - see the LICENSE file for details.

AI Dev Toolkit - MCP Server