哈喽我是远景,一个天天跟 AI Agent 打交道的程序员。
不知道你们有没有这种痛点:平时干活全靠 Claude Code,但它只能在电脑上用,出门在外、电脑不在身边的时候,突然想改个代码、跑个脚本、查个项目文件,根本没法弄,干着急。前阵子 OpenClaw、Hermes 这些个人 AI 助理火的时候,我也挨个试了,新鲜劲过了就全吃灰了 —— 毕竟真要处理重任务、写代码,还是得靠 Claude Code 这种正经生产力工具,花里胡哨的功能没啥用。

直到我最近挖到了逛逛 GitHub开源的wechat-claude-code,用了快俩月,真的太香了,完全解决了我出门用不了 Claude Code 的痛点,今天必须给你们好好安利一下。

01项目简介
说白了就是个 Skill,能把你的微信和电脑上跑的 Claude Code 完全打通。扫码绑定之后,你微信里会多一个好友,你给它发的所有消息、图片、语音、文件,都会自动转给你电脑上的 Claude Code,它的回复、生成的文件也会实时推回微信 —— 就跟跟真人聊天一样自然。不用开远程桌面,不用搭复杂的服务,装完就能用,体验特别顺。
开源地址先放这了,感兴趣的可以先去 star 一波:
👉 https://github.com/Wechat-ggGitHub/wechat-claude-code

开源地址:https://github.com/Wechat-ggGitHub/wechat-claude-code
02 可以干啥?
这个项目最近迭代特别快,更了好几个版本,我一路用过来,每个新功能都精准戳中了平时用的痛点,给你们说说我觉得最爽的几个点:
1. 微信消息终于不刷屏了,只给你看有用的
最早的版本我也试过,那会 Claude Code 干啥都往微信推,各种中间调试信息、过程日志全发过来,手机叮叮叮响个不停,翻半天才能找到有用的内容,跟看日志似的。
最近的版本专门做了信息过滤,只把核心的进度、最终结果推过来,干净利落,有用但不繁杂,再也不用在一堆消息里扒重点了。

2. 长时间没反应?给你发个安抚,知道它没挂
用 Claude Code 跑大任务的时候,比如改整个项目的代码、生成大文件,经常十几分钟没动静,你也不知道它是卡死了还是在干活,隔一会就得跑回电脑跟前看一眼,特别闹心。
作者专门加了个特别有意思的小功能:如果超过 5 分钟没响应,就会从安抚话术池里随机挑一句发给你,比如「正在努力干活中,稍等哦~」「这个任务有点复杂,再给我两分钟」。
说真的,这个不起眼的小功能是我平时用的最爽的,不用一直悬着心等,知道它还在跑就行,体验感直接拉满。

3. 加上「对方正在输入中」,体验更像真聊天
之前一直没做这个,最近的版本也加上了。只要 Claude Code 在处理任务,微信聊天顶部就会显示「对方正在输入中...」,跟跟真人聊天的感觉一模一样,不会你发了消息之后一片空白,不知道啥情况,体验顺了特别多。
4. 手机电脑体验完全一致,不会像两个 AI
这个是我最在意的一点,也是我之前 pass 掉好多同类工具的原因:
很多微信接入 Claude Code 的插件,电脑上是一个样,手机上发过去的指令,出来的结果完全不一样,就像用了两个不同的 AI,体验割裂的要死。原因也很简单,它们走的是 SDK 模式,不是 CLI 模式,底层逻辑根本不一样。
这个项目就把这个问题解决了:现在你在电脑上输一个指令,Claude Code 怎么编排、怎么输出,你在手机上发一模一样的指令,效果几乎完全一致,就是同一个 AI 的感觉。而且现在默认就是全自动模式,不用你反复确认,跟电脑上用的体验一模一样,这点真的太重要了。

5. 全格式文件收发,电脑不在身边也能拿文件
这个功能我现在天天用,救过我好多次急:
你不光能给它发图片、语音、Word、PDF 各种文件,它生成的文件也会直接通过微信发给你。比如你让它写篇文章、改个脚本,不会只跟你说「我写完了,放桌面了」,而是直接把文件给你推到微信里,点开就能看。
更爽的是,你电脑不在手边的时候,直接跟它说「把我电脑上某某项目的 README 文件发给我」,它直接就给你发过来了,不用再麻烦同事或者跑回家开电脑,出门在外查个文件太方便了。
想试的朋友特别简单,不用你手动装依赖配环境,全让 Claude Code 自己搞定:直接把这句话发给你的 Claude Code:
方式一:skills CLI(推荐)
npx skills add Wechat-ggGitHub/wechat-claude-code
首次在对话中触发时,会自动克隆项目源码并安装依赖。
方式二:手动克隆
git clone https://github.com/Wechat-ggGitHub/wechat-claude-code.git ~/.claude/skills/wechat-claude-codecd ~/.claude/skills/wechat-claude-code && npm install
/wechat-claude-code,就会弹出来一个二维码,拿微信扫一下就能用了。后续也可以用 /wechat-claude-code命令来管理进程,启动、重启、停止都可以,特别方便。

03 总结
用的时候发现哪里不爽,直接在微信里给 Claude Code 说,让它分析问题、给解法、检查完就修复、重启 Skill,体验没问题就直接推到 GitHub 上。
一边用一边改,哪里体验不好改哪里,迭代速度特别快 —— 这不就是 AI 编程最爽的地方吗?
Hey! I'm Yuanying, a programmer who works with AI Agents every day.
Have you ever felt this pain: you rely on Claude Code for your daily work, but it can only be used on your computer? When you're out and about or your computer isn't available, needing to edit code, run a script, or check project files is impossible — you're stuck. When personal AI assistants like OpenClaw and Hermes became popular, I tried them all, but after the initial novelty wore off, they were collecting dust. For serious tasks like writing code, I still needed a proper productivity tool like Claude Code; fancy features weren't helpful.
Then, recently, I discovered a GitHub project called wechat-claude-code developed by Guangguang, and I've been using it for nearly two months. It's amazing — it completely solves the problem of not being able to use Claude Code while away. Today, I want to share this with you all.
01 Project Overview
Essentially, this is a Skill that fully integrates your WeChat with the Claude Code running on your computer. After scanning the code to bind them, you'll have a new friend in WeChat. Everything you send to this friend — messages, images, voice notes, files — is automatically forwarded to your Claude Code, and its responses and generated files are sent back to WeChat in real time. It's as natural as chatting with a person, without the need for remote desktop or complex services; install it and you're ready to go.
Here's the open-source address; feel free to check it out and star it:
👉 https://github.com/Wechat-ggGitHub/wechat-claude-code
02 What Can It Do?
This project has been iterating very quickly, with several new versions. I've been using it throughout, and each new feature precisely addresses my pain points. Let me share the highlights:
1. WeChat messages are no longer spammy — only useful information is shown.
In the early versions, Claude Code would send everything to WeChat — intermediate debugging info, process logs — flooding my phone with notifications. It was like reading logs, searching for what mattered. Now, with the new version, information filtering is implemented: only key progress updates and final results are pushed, making it clean and concise.
When running large tasks (like modifying entire projects or generating big files), Claude Code can be inactive for 10–15 minutes. You wouldn't know if it was stuck or working, so you'd have to check the computer repeatedly. The author added a clever feature: if there's no response for over 5 minutes, a random encouraging message is sent, like "Working hard on it, hang tight~" or "This task is complex, give me a few more minutes." This small feature is incredibly satisfying — no more anxiety, just peace of mind that it's still running.
3. "Typing..." indicator added for a more chat-like experience.
Previously, once a message was sent, the chat would be silent until the response came. Now, while Claude Code is processing, the WeChat header displays "The other party is typing...", mimicking real human conversation and reducing uncertainty.
4. Consistent experience between phone and computer — not two different AIs.
5. Full format file transfer — access files even when away.
This feature saves me every time. You can send images, voice notes, Word documents, PDFs, etc., to the bot, and it will send back its generated files (like reports or scripts) directly to WeChat. More importantly, if you're away from your computer, you can ask it to send specific files (e.g., "Send me the README of project X"), and it will comply — eliminating the need to return to the office or wait for your return.
Getting started is very easy — no need to manually install dependencies or configure environments. Let Claude Code do it all: simply send the following command to your Claude Code instance:
The first time this is triggered in a conversation, it will automatically clone the project source code and install dependencies.
Method two: manual clone
Once the installation and deployment are complete, restart Claude Code and type /wechat-claude-code. A QR code will appear — scan it with WeChat to bind the account. You can also use this command to manage the service (start, restart, stop) later on.
03 Conclusion
The cool thing about this project is that it's truly built by users for users. The author, Guangguang, is very active — you can often find him on the project's GitHub or the QQ group. If you encounter any issues while using it, you can report them directly to the bot in WeChat, which will analyze the problem and suggest solutions. The iterative speed is very fast, continuously improving the user experience.
方式二:手动克隆
Wait until the installation and deployment are complete, then restart Claude Code. Input /wechat-claude-code, and a QR code will appear. Scan it with WeChat to use. You can also use the /wechat-claude-code command to manage the process, allowing you to start, restart, and stop it easily.
03 总结
If you find something unsatisfactory during use, simply mention it to Claude Code in WeChat. It will analyze the problem, provide a solution, perform checks to fix it, and restart the Skill. Once the experience is verified, you can push it directly to GitHub.
You can make changes while using the service — improve any part that has a poor experience. This rapid iteration is exactly what makes AI programming so exciting.

