llm-coding-agent 0.1a0
Simon Willison released llm-coding-agent 0.1a0, an experimental coding agent built on his LLM library. The agent provides tools for reading, editing, searching files and executing commands, and ships with a Python API and CLI. It was developed using Claude Code (Fable 5) via TDD with a spec-first approach, and is available as a slop-alpha on PyPI.
Background
- Simon Willison is a prominent Python developer who created **LLM** (a CLI tool for chatting with AI models) and **Datasette** (an open-source data exploration tool).
- **LLM** has evolved into an "agent framework" — it can now orchestrate multi-step tasks where the AI reads/writes files, runs commands, and calls tools autonomously.
- **Claude Code** (by Anthropic) is a popular coding agent that edits files, runs shell commands, and manages git — Willison set out to build an open-source equivalent on top of LLM.
- **Fable 5** is an alternate name for Claude 3.5 (part of a renaming scheme Willison and others adopted after Anthropic deprecated older model names).
- This entire project was written by AI: Willison gave Claude Code two prompts — one to write a spec, one to build the project using TDD — and shipped the result.
- **uvx** runs Python packages from PyPI without installing; `--yolo` skips all approval prompts, letting the AI act without human confirmation.
- The resulting `llm code` agent gives the LLM tools (read/edit/write files, run commands, search code) and iterates until a task is done.