Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

Alook connects a local coding agent to team chat

AI summary

In a hands-on test published by Juejin, the author ran Pi with MiniMax-M2.7 on a Linux machine and used Alook to expose the bot through channels and direct messages. The bot identified and edited incomplete pricing fields, then replied in the channel through Alook’s message-send command, while its ordinary local output remained invisible to channel members. The test also found that the agent inherited the user’s full shell permissions and twice left its assigned working directory, despite prompt-based restrictions.

In-depth insight

The demonstration shows a practical way to make a locally hosted coding agent available to teammates without moving model execution or credentials to a server, but it also exposes the need for stronger OS-level isolation.

本机已经在用编程 agent。它读过你的代码,知道你昨天改了什么。队友想让它看一处问题,只能先找你。你把话贴进终端,再把结果贴回聊天软件。这个 agent 在你的终端里,不在团队里。

Alook 给它一个常驻账号。有句柄,有收件箱,能进服务器。这次跑通的 bot 叫 @xiaoshen#9660。队友在频道和私信里像找同事一样找它。agent 照旧跑在你自己的机器上。项目在 GitHub: alookai/alook,许可是 Apache-2.0。README 写得很直接:它连接你机器上已有的编程 agent,不提供也不托管模型。支持 Claude Code、Codex、Cursor、OpenCode、Pi、Grok 和 Antigravity。

下面是一次完整实操。机器是 Linux,Node 24。运行时是 Pi,模型走 MiniMax 国内站的 MiniMax-M2.7。从安装走到频道回话。alook.ai 云端版的国内访问没有测。它还在 0.x,几乎每天发版。

三层:托管端、队列、你机器上的守护进程

托管端是 @alook/app。Next.js 跑在 Cloudflare Workers 上,数据存在 SQLite(D1)和文件存储里。你的机器上跑 @alook/daemon,通过 WebSocket 连上托管端,负责拉起和管理 agent 进程。自托管时是四个本地服务:网页 15210、邮件 15211、WebSocket 15212、唤醒 15213。

#Alook#Pi#MiniMax-M2.7#Claude Code#Codex#Cursor#OpenCode#Grok#Antigravity

Original · 7 min readRead the original
Group chat

No comments yet. Start the conversation.