A version of a command-line interface (CLI) that leverages Claude models to create and manage its own tools dynamically during conversations. This framework allows the AI to autonomously expand its capabilities in real-time, becoming more effective with continued use, targeting developers and researchers interested in advanced AI agent functionality.

How It Works

The core of Claude Engineer is a dynamic tool ecosystem where the AI autonomously creates, manages, and executes its own tools. This approach enables continuous capability expansion during interactions, adapting to user needs through smart dependency management and automated tool code generation. It integrates with any Claude model, tracks token usage, and refines existing tools with enhanced error handling and a dynamic module importing system.

Quick Start & Requirements

• Installation: Clone the repository, navigate to the directory, create and activate a Python virtual environment (venv), then install dependencies using pip install -r requirements.txt or pip install -e .. Run the agent via python main.py (Windows) or python3 main.py (macOS/Linux). The REPL can also be launched with python -m claude_engineer.
• Requirements: Windows or macOS, Python 3.6+, an Anthropic API Key.
• Configuration: Set ANTHROPIC_API_KEY in a .env file. Optional configurations include CE_MODEL, CE_MAX_TOKENS, CE_VERBOSE, and CE_TOOLS_DIR.
• Docs/Demo: Type /help within the REPL for commands.

Highlighted Details

• Dynamic tool creation and autonomous management by the AI.
• Continuous expansion of capabilities during conversations.
• Integration with any Claude model (e.g., Opus 3-4.7).
• Built-in tools include file system operations (read_file, write_file, list_dir), shell command execution (run_shell with 60s timeout), and web search (web_search).
• Real-time display of tool execution and token usage tracking via Anthropic API.

Maintenance & Community

Contributions are welcome; users are advised to open an issue for non-trivial changes before submitting a Pull Request. No specific community links (e.g., Discord, Slack) or roadmap details are provided in the README.

Licensing & Compatibility

The project is released under the MIT License, allowing free use, modification, and distribution, with a request for contributions back. This license is generally permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

Requires an Anthropic API key, which incurs costs. The agent's effectiveness is tied to the capabilities of the underlying Claude model and the quality of generated tools. Specific performance benchmarks or detailed error handling scenarios are not provided.