开源 runmarq:在托盘估时并通知编码智能体何时结束
runmarq v0.1.0 在系统托盘为 Claude Code、Codex、Cursor 提供本地估时与完成/失败通知,无云端、需源码构建。
开发者 arjunpundeer15 发布开源工具 runmarq v0.1.0:在系统托盘里估计 AI 编码智能体还要多久结束当前任务,并在完成、失败或需要人工介入时通知。
作者称,随着编码智能体承接更长任务,自己反复盯终端进度,于是做了这个本地工具。当前支持 Claude Code、Codex 与 Cursor;根据本机历史运行学习估时;不发起 AI 调用、不要求账号、不上传数据。技术栈为 Rust、Tauri、React 与 Next.js。项目公开在 GitHub,网站为 runmarq.site,目前需从源码构建,欢迎反馈。
对长时间跑智能体的个人开发者,这类「进度与注意力」工具补的是 IDE 之外的空档:不必为了确认是否卡住而反复切窗。早期版本功能边界仍应以仓库说明为准,估时精度取决于本机历史样本是否足够。若你同时开多个智能体会话,托盘汇总完成与失败状态,有助于把注意力放回真正需要人工介入的步骤,而不是盯着滚动日志。安装与权限方面,因需访问本机进程与历史记录,建议只在受信任环境使用,并核对仓库最新发行说明。
runmarq v0.1.0 open-sourced for coding agent ETA
Built runmarq, an open-source tool that estimates how long your AI coding agent will take to finish a task.
As coding agents take on longer tasks, I found myself repeatedly checking the terminal for progress. runmarq is my attempt to solve that.
It runs quietly in your system tray, providing live time estimates and notifications when an agent finishes, fails, or needs attention.
- Supports Claude Code, Codex, and Cursor
- Learns from your local run history
- No AI calls, accounts, or data uploads
- Built with Rust, Tauri, React, and Next.js
v0.1.0 is now public on GitHub. It's an early release, available to build from source, and feedback is welcome.
Website: https://t.co/JckmZhkjTG
GitHub: https://t.co/OWxvDTp5Xc