linearuncle/xharness
A macOS AI coding buddy that reads, edits, and runs your project's code directly on your own machine
xharness is a macOS desktop app (plus a CLI) where you point it at a project folder and give it tasks in chat, and it reads code, edits files, and runs commands to verify its work, all visibly step by step. It plugs into DeepSeek, Kimi, Anthropic-compatible models, and more, with all data stored locally in a SQLite file. A built-in guardrail called AgentGuard intercepts risky delete-type commands before they run and asks for your confirmation.
What it does
- Once you register a project folder, the AI reads and edits files and runs terminal commands directly, showing in real time which files it read, which commands it ran, and its reasoning steps
- You can paste a screenshot straight into the chat to have it build a page from a design mockup, use @ to search and reference project files, and use / to invoke reusable preset instructions called skills
- DeepSeek is built in by default, and you can add and switch between other services like Kimi, the official Anthropic API, or xAI's Grok that use either the Anthropic Messages or OpenAI Responses API formats
- xharness has only one operating mode, full access, where the AI reads, writes, and executes commands without asking permission first, but its built-in AgentGuard plugin intercepts dangerous delete operations like rm, git clean, or SQL DROP/DELETE and hands the decision back to the user
- All data, including chat history and API keys, is stored in a single local SQLite file on your Mac with no cloud sync
Why it matters
Rather than a copy-paste chatbot, this lets an AI actually work inside your project while you watch every step, which could change day-to-day coding workflows. But because it runs in a no-permission full-access mode, the project itself warns against pointing it at directories without backups or machines holding production credentials.
Terms in this repo
- AgentGuard · xharness's built-in safety plugin that intercepts risky delete commands before execution and asks the user to confirm
- Full access mode · xharness's only operating mode, where the AI reads, writes files, and runs commands without asking for prior approval
- Skill · A reusable instruction template you save in advance and invoke with a slash command
- Anthropic Messages / OpenAI Responses · The two AI model API communication formats xharness supports
Repository description (English)
🤝 装在你电脑上的 AI 编程搭档（macOS GUI + CLI）：把项目文件夹丢给它，用聊天布置任务——自动读代码、改代码、跑命令验证，全程可见。支持 DeepSeek / Kimi / 任意 Anthropic 兼容模型；截图粘贴看图干活；技能系统；数据全本地。
Open on GitHubTrending repos
- vorssaint/vorssaint-utilsOne free menu bar app replaces a dozen paid Mac utilities
- Alishahryar1/free-claude-codeA local proxy that lets coding AI agents run on 49 free or cheap model providers instead of one paid service
- freestylefly/awesome-gpt-image-2A library of 532 reverse-engineered prompts that turn GPT-Image2 into a predictable image-making tool
- block/buzzAn open-source workspace where humans and AI agents chat, code, and review in the same rooms
- NousResearch/hermes-agentNous Research's Hermes is an AI agent that gets smarter the more you use it
- virgiliojr94/book-to-skillA tool that turns technical book PDFs into on-demand reference skills for AI coding agents
- VoltAgent/awesome-agent-skillsA single hub collecting over 1000 'how-to' manuals that make AI coding assistants act like experts
- anthropics/claude-plugins-communityA shared shelf where anyone's Claude add-ons get listed for install