he-yufeng/CoreCoder

Health v2 · maintenance only

Flags: young

Adoption not part of the score

1678 stars · 395 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

CoreCoder is a minimal AI coding agent written in roughly 1,000 lines of Python, inspired by Claude Code and designed to work with any LLM. It is intentionally readable end to end so developers can understand how coding agents work and fork it to build their own.

Use cases

• understand how an AI coding agent works internally
• learn the core agent loop and tool-calling pattern
• fork a minimal coding agent to build your own
• run an LLM-powered coding assistant in the terminal
• study a readable alternative to Claude Code or aider
• teach coding agent architecture with a small codebase

When to choose

• you want to read and learn from a tiny, complete coding agent implementation
• you need a hackable base to fork into your own agent
• you want a minimal CLI coding agent that works with any LLM provider

When to avoid

• you need a production-grade coding assistant with extensive features
• you want polished terminal pair-programming like aider
• you expect the robustness and safety of a mature commercial tool

Facets

cli-tool · maturity active

agent-framework llm-inference cli developer-tools developer-tools large-language-models education python cli cross-platform coding-agent claude-code-inspired minimal learning-resource readable-source nanogpt-for-agents ai-agents command-line

Member repositories

For agents

markdown · JSON · MCP: product_card(name="he-yufeng/CoreCoder")