Simon Willison: llm-coding-agent 0.1a0 — A Python Library Implementing a Claude Code-Style Coding Agent

Main idea

Simon Willison released llm-coding-agent (version 0.1a0), an open-source Python library implementing an agentic coding loop in the style of Claude Code. The library provides tools for reading and editing files and executing shell commands — the core building blocks of AI coding agents.

Context

Willison has long experimented with LLM tools and builds on top of his own llm Python library. The release comes as AI coding agents (Claude Code, Codex, Cursor) dominate discussion — but most are proprietary. Willison offers an open-source alternative as a proof-of-concept and educational reference implementation.

Why it matters

For developers, this is a practical demonstration of how to build your own AI coding agent from scratch. The code is readable, minimal, and well-commented — an ideal foundation for those wanting to understand the internal mechanics of tools like Claude Code.

Details / arguments

• Version: 0.1a0 (alpha) — active development
• Tools: file reading, file editing, shell command execution
• Demo: agent creates a SwiftUI CLI app for displaying ASCII art time
• Built on Willison's llm Python library (model-agnostic)
• Open-source, available on PyPI