shareAI-lab/learn-claude-code

This page introduces an external open-source repository. It is not an HDATF product.

What it does

A Python tutorial that builds a small Claude Code style agent harness from scratch, on the idea that bash is all you need. It argues that agency comes from the model and the harness lets it act in an environment.

How it helps ATF

A reference for Harness, which is also an execution layer around a model. Its small step-by-step build is worth comparing when explaining the parts of a harness to new engineers.

License

MIT Permissive. Commercial use and changes are allowed if the copyright notice is kept.

More in this category

• code-yeongyu/oh-my-openagent
  A TypeScript orchestration project for AI coding agents that is started by adding a keyword to a prompt. The README excerpt mentions multi-model coding work and a memory system, but gives few other details.
• x1xhlol/system-prompts-and-models-of-ai-tools
  A collection of system prompts, internal tool definitions and AI model information gathered from many AI products, including Claude Code, Cursor, Devin AI, Manus, NotionAI, Perplexity and v0.
• daytonaio/daytona
  Daytona is a runtime that runs AI-generated code and agent workflows in isolated sandboxes, each with its own kernel, filesystem and network stack, managed through SDKs, an API and a CLI. This public repository is no longer maintained.
• bytedance/UI-TARS-desktop
  A multimodal agent stack with two projects. Agent TARS brings GUI agent and vision abilities to the terminal, computer and browser with MCP tools, and UI-TARS Desktop operates local or remote computers and browsers using the UI-TARS model.
• MiroMindAI/MiroThinker
  MiroThinker is a deep research agent, with released models, built for complex research and prediction tasks. The project includes benchmark evaluation and trace collection, and its online version generates research reports that can be previewed and shared.

Only repositories in the ranked Trendshift lists are included, and the lists are used only to find candidates. We do not copy their ranks. Descriptions, licenses and star counts come from each GitHub repository. The notes are our own reading. We have not tested these projects, and a place on a trending list does not prove quality.