AI UNDERDOGSDAILY PICK
AI UNDERDOGS
并行跑 Claude Code，终于不瞎了
Stop guessing when running Claude Code in parallel
neochoon/agenthud
同时开好几个 Claude Code
哪个卡了、哪个跑完了
全靠猜——直到我发现了这个
Run a few Claude Code sessions at once
and it's pure guesswork which one's stuck or
done — until I found this
TUI 面板，不开浏览器
TUI dashboard, no browser needed
它叫 AgentHUD
是专门给 Claude Code 用户做的终端监控面板
用 Claude Code 的人本来就活在终端里
这个工具直接在那里给你一个 HUD——左边是会话列表
中间是实时输出
右边是 sub-agent 的树状结构
AgentHUD is a terminal-native dashboard built for Claude
Code users. Since you already live in the
terminal, it puts the HUD right there —
session list on the left, live output in
the center, sub-agent tree on the right
★ SIGNAL 1
657 次提交，不是周末项目
657 commits — not a weekend project
点进 repo 一看
657 次提交、72 个 tag
22 个分支，CI 配好了
demo 录屏也有
这个项目的工程成熟度不是业余水平
Open the repo: 657 commits, 72 tags, 22
branches, CI set up, demo recordings included. This
thing is built to production-grade standards
★ SIGNAL 2
让 AI 总结你的 AI 做了啥
AI that summarizes what your AI agents did
最有意思的是这个功能：它会用 LLM 自动生成日报和周报
总结你各个 Agent 的工作进展
用 AI 监控 AI
再让 AI 写总结——这层 meta 玩得挺妙
The fun part: it uses an LLM to
auto-generate daily and weekly digests summarizing your agents'
progress. AI monitoring AI, then writing summaries about
it — that's a nice meta layer
作者自己说得很直接：这就是给 Claude Code 做的实时监控面板
支持并行会话和 sub-agent 追踪
外加 LLM 生成的摘要
没有花里胡哨的营销
repo 本身就是最好的说明
The author's description is dead simple: a live
monitor for parallel sessions and sub-agents, with LLM-generated
digest. No hype — the repo speaks for
itself
AI UNDERDOGS
给你的 Claude Code 加个 HUD
Give your Claude Code sessions a HUD
neochoon/agenthud
关注 · 每天发现更多 AI 神作
github.com/neochoon/agenthud