agentforce314/clawcodex ? reverse-engineered prompt

Build me a Python based coding agent that feels like Claude Code, but works as a real product I can run from the terminal.

I want to be able to start it in any project, log in with a model provider, and then chat with it to inspect files, edit code, run shell commands, and keep going across sessions. It should have a simple CLI, a good terminal UI, and a basic web view too if that’s not too much extra. Make the setup easy, with a first run login flow, config saved locally, and clear commands for things like permissions, settings, update, and health checks.

It should be careful about cost and token usage, so if you can add a compact mode for command output and keep the prompt history efficient, that would be great. If you need to check current docs for provider setup or packaging details, look them up online. Keep it production ready, reliable, and easy to use without me having to think about the internals.