AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
翻译准备中
構建 coding-agent 工作的工作區。
AI 翻译
Build the workspace your coding agents work in.
取自仓库 README
正在加载完整 README。 以下为已保存的摘录。
Build the workspace your agents work in. The screen is yours to draw; we keep the agents, terminals and diffs alive inside it.
Four agents, four terminals, four tabs. One is waiting for approval, one finished ten
minutes ago, one is editing a file you meant to read first — and you find out by cycling
through windows.
One screen instead. Each agent's terminal, the diff it is editing, the issues, a browser —
grouped into frames bound to a repo, a worktree or an SSH host.查看完整 README ↗Everything below was written by AI. This line wasn't. — dip497
[!NOTE] Yes, it's Electron. Please don't kill me. 🙏 I'm a backend guy, Hate mail goes to
/dev/null.
Five agents in five terminal tabs?
Hivemind puts every coding agent on one board, each in a real terminal: Claude Code, Codex, Pi, and whatever ships next week. Then it gets weird, in a good way. Ask an agent for a view, a widget or a tool, and it builds one and plugs it in while you watch.
hive ctl an agent starts, reads and steers other agents.
Your scripts can too, through the API.More weird stuff on HiveHub, or ask your agent to make some.
Linux x86_64 and macOS on Apple Silicon:
bash <(curl -fsSL https://hivemind.griiken.com/install.sh)
Windows x64, in PowerShell:
irm https://hivemind.griiken.com/install.ps1 | iex
Other ways to install, and what it puts where: https://hivemind.griiken.com/guide/install/.