Hanzo REPL

An intimate REPL environment for testing Hanzo's Model Context Protocol (MCP) tools and AI integration. Think of it as rlwrap meets Claude Code - direct access to MCP tools with IPython magic.

Features

• 🎯 Direct MCP Access: All MCP tools available as Python functions
• 💬 Integrated Chat: Chat with AI that can use MCP tools
• 🔧 IPython Magic: Use ? and ! helpers, tab completion
• 🔄 Live Editing: Edit the REPL source code on the fly
• 🧪 Built-in Tests: Comprehensive test suite for MCP tools
• 🎨 Rich Output: Beautiful formatting with Rich library

Quick Start

# Setup (one time)
make setup

# Start the REPL
make dev

Usage Examples

Direct Tool Access

# All MCP tools are available as functions
>>> read_file(file_path="/etc/hosts")
>>> write_file(file_path="test.txt", content="Hello, World!")
>>> search(query="def main", path=".")
>>> run_command(command="ls -la")

Chat with AI

# Simple chat
>>> chat("What files are in the current directory?")

# AI will use MCP tools to answer
>>> chat("Create a Python script that prints the current time")

# Complex workflows
>>> chat("Find all Python files with 'test' in the name and show their sizes")

IPython Magic Commands

# Single-line chat
%chat What is the weather today?

# Multi-line chat
%%ai
Can you help me create a web scraper?
I need it to extract titles from a list of URLs.

# List available tools
%tools

# Execute specific tool
%tool read_file {"file_path": "README.md"}

# Edit REPL source
%edit_self ipython_repl.py

# Change model
%model claude-3-opus-20240229

Object Access

# Access MCP server directly
>>> mcp.tools.keys()

# Access LLM client
>>> llm.get_available_providers()
>>> llm.set_model("gpt-4")

# Access tool executor
>>> executor.get_context()

Available Commands

Makefile Targets

• make dev - Start IPython REPL with MCP tools
• make test - Run test suite
• make demo-file - Demo file operations
• make demo-chat - Demo chat functionality
• make demo-search - Demo search operations
• make clean - Clean generated files

REPL Commands

• chat(message) - Chat with AI using MCP tools
• tools.<tab> - Tab completion for all tools
• %tools - List all available MCP tools
• %model - Show/set current LLM model
• %edit_self - Edit REPL source code

Environment Variables

Set at least one LLM provider API key:

• OPENAI_API_KEY - OpenAI API key
• ANTHROPIC_API_KEY - Anthropic/Claude API key
• GROQ_API_KEY - Groq API key
• Other providers supported by litellm

Integration with Hanzo Chat

The REPL can be used as an MCP server for Hanzo Chat:

1. Start the REPL: make dev
2. Configure Hanzo Chat to use the MCP endpoint
3. All tools become available in the chat interface

Testing

# Run all tests
make test

# Run integration tests (requires API keys)
make test-integration

# Interactive test in REPL
make test-repl

Development

# Watch for changes and auto-restart
make watch

# Format code
make format

# Lint code
make lint

# Type check
make type-check

Architecture

hanzo_repl/
├── ipython_repl.py    # IPython-based REPL with magic commands
├── llm_client.py      # Multi-provider LLM client
├── tool_executor.py   # Execute tools based on LLM responses
├── tests.py          # Comprehensive test suite
└── cli.py            # Command-line interface

mcp_repl.py           # Minimal direct-access REPL

Why This Exists

This REPL provides a more intimate way to interact with Hanzo's MCP tools:

• No barriers between you and the tools
• Direct Python access to all functionality
• Chat with AI that can modify its own code
• Perfect for testing and development
• Great for learning the MCP API

Think of it as your personal MCP playground where you can experiment, test, and build with immediate feedback.