// the find
Doriandarko/claude-engineer
Claude Engineer is an interactive command-line interface (CLI) that leverages the power of Anthropic's Claude-3.5-Sonnet model to assist with software development tasks.This framework enables Claude to generate and manage its own tools, continuously expanding its capabilities through conversation. Available both as a CLI and a modern web interface
Claude Engineer is a Python CLI (and optional Flask web UI) that wraps Claude 3.5 Sonnet into a self-modifying coding assistant — it can write new tool files to disk at runtime and load them into the running session. It's aimed at developers who want an agentic coding companion that expands its own capabilities through conversation, rather than being limited to a fixed tool set.
The self-modifying tool architecture is genuinely interesting: Claude writes Python files into the tools/ directory, and the framework hot-reloads them without restarting. Token tracking is done properly using Anthropic's token counting API rather than estimating, and the context window management is explicit rather than silently truncating. The sandboxed code execution via E2B is the right call — letting an LLM run arbitrary code locally would be a bad default. The dual interface (CLI via Rich + Flask web UI) means you can actually choose based on context rather than being forced into a browser.
Last commit was December 2024 and it's hardcoded to claude-3-5-sonnet — there's no obvious path to swap the model, which means it's already behind. The self-improving tool creation is a compelling demo, but tools written by the LLM and saved to disk with no review step is a real security surface: a malicious prompt could cause Claude to write and execute arbitrary code. The Flask web server has no authentication whatsoever, so running app.py on anything but localhost exposes a code execution endpoint to the network. There's also no persistence layer — conversation history lives in memory, restart and it's gone.