SawyerHood/gitclaw

Registry 信息

项目简介

GitClaw 是一个基于 Python 的开源工具，完全在 GitHub Actions 上运行，无需外部基础设施即可实现自动化代理工作流。

要点

• 完全在 GitHub Actions 上运行
• 基于 OpenClaw 框架构建
• 自动化代理执行
• 无需外部基础设施
• 基于 Python 实现

使用场景

• 自动化 CI/CD 代理任务
• 定时代理工作流
• 轻量级代理部署
• GitHub 原生自动化

README 摘要

 A personal AI assistant that runs entirely through GitHub Issues and Actions. Like [OpenClaw](https://github.com/openclaw/openclaw), but no servers or extra infrastructure. Powered by the [pi coding agent](https://github.com/badlogic/pi-mono). Every issue becomes a chat thread with an AI agent. Conversation history is committed to git, giving the agent long-term memory across sessions. It can search prior context, edit or summarize past conversations, and all changes are versioned. Since the agent can read and write files, you can build an evolving software project that updates itself as you open issues. Try asking it to set up a GitHub Pages site, then iterate on it issue by issue. ## How it works 1. **Create an issue** → the agent processes your request and replies as a comment. 2. **Comment on the issue** → the agent resumes the same session with full prior context. 3. **Everything is committed** → sessions and changes are pushed to the repo after every turn. The agent reacts with 👀 while working and removes it when done. ### Repo as storage All state lives in the repo: ``` state/ issues/ 1.json # maps issue #1 -> its session file sessions/ 2026-02-04T..._abc123.jsonl # full conversation for issue #1 ``` Since sessions are in git, the agent can grep its own history and edit or summarize past conversations. ## Setup 1. **Fork this repo** 2. **Add your Anthropic API key** - go to **Settings → Secrets and variables → Actions** and create a secret named `ANTHROPIC_API_KEY`. 3. **Open an issue** - the agent starts automatically. 4. **Comment on the issue** - the agent resumes where it left off. ## Security The workflow only responds to repository **owners, members, and collaborators**. Random users canno

话题