Create a CLI coding agent from a Python LLM base

Claude Code turned Simon Willison’s Python LLM template into a working CLI coding agent from two prompts.

Done with Custom agent· Advanced

What they did
Simon Willison started a fresh repo from his Python library template and handed Claude Code for web two tasks. The first asked for a spec.md describing a Claude Code-style agent for the llm alpha; the second asked for an implementation plan using red/green TDD, sensible commits, docs, and manual testing with the OpenAI key available in the environment.
How it went
The agent reached an alpha release on PyPI, and Willison said it looked pretty good for a first try. It gained tools for reading and changing files plus running commands, and it also exposed a Python API he had not requested. In a GPT-5.5 demo, it built a clock display made from text characters.
Tips
The demo relied on --yolo, so approvals were skipped entirely. The README also mentions narrower allow-list options such as --allow pytest* if you want tighter control.

Try it yourself with Claude Code

Start from [Python LLM library or template] and build a CLI coding agent for [project name]. It should be able to inspect files, edit files, and execute shell commands in the current directory. First write a short spec, then implement it with small test-driven steps, clear commits, and a README. Before any step that would send data, overwrite work, delete files, or submit something externally, pause and ask me to approve it. When the build is done, run it on a simple demo task in [scratch directory] and print the transcript.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on simonwillison.net ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.