AI Skill Hub 推荐使用:开源AI工作流 是一款优质的Agent工作流。AI 综合评分 7.5 分,在同类工具中表现稳健。如果你正在寻找可靠的Agent工作流解决方案,这是一个值得深入了解的选择。
Multi-agent workspace protocol - context sharing, event bus, and coordination fo,提供AI工作流管理和协调功能,帮助开发者高效管理项目。
开源AI工作流 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
Multi-agent workspace protocol - context sharing, event bus, and coordination fo,提供AI工作流管理和协调功能,帮助开发者高效管理项目。
开源AI工作流 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一:npm 全局安装 npm install -g ufoo # 方式二:npx 直接运行(无需安装) npx ufoo --help # 方式三:项目依赖安装 npm install ufoo # 方式四:从源码运行 git clone https://github.com/Icyoung/ufoo cd ufoo npm install npm start
# 命令行使用
ufoo --help
# 基本用法
ufoo [options] <input>
# Node.js 代码中使用
const ufoo = require('ufoo');
const result = await ufoo.run(options);
console.log(result);
# ufoo 配置说明 # 查看配置选项 ufoo --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量(覆盖配置文件) export UFOO_CONFIG="/path/to/config.yml"
ufoo is a project-scoped multi-agent workspace runtime. It wraps Claude Code, OpenAI Codex, Antigravity, and ufoo's native ucode agent with a shared chat dashboard, daemon, event bus, memory, reports, group orchestration, and terminal launch modes.
Package: u-foo on npm
- One TUI dashboard for launching, watching, messaging, and resuming agents. - Project daemon over .ufoo/run/ufoo.sock for launch/resume, reports, groups, cron, and controller routing. - Project-local event bus for agent-to-agent messages, wakeups, queue checks, and activation. - Shared context primitives: decisions, durable memory, prompt history, reports, and agent registry state. - Launch modes for internal, tmux, host, Terminal.app, and iTerm2 workflows. - Built-in group templates for launching and orchestrating multi-agent workflows. - ucode, a native ufoo coding-agent runtime. - ufoo mcp, a local global MCP bridge for external MCP-capable agents.
- Node.js 18 or newer. - macOS for Terminal.app/iTerm2 integration. - Claude Code, Codex CLI, or Antigravity CLI installed when using the matching wrappers: uclaude, ucodex, or uagy.
Install the published package:
npm install -g u-foo
Or link this repository for local development:
git clone https://github.com/Icyoung/ufoo.git
cd ufoo
npm install
npm link
Installed binaries:
| Binary | Purpose |
|---|---|
ufoo | Main CLI, chat dashboard, daemon, local global MCP bridge, groups, bus, context, memory, reports, and online helpers. |
uclaude | Claude Code wrapper with ufoo bootstrap and bus identity. |
ucodex | Codex wrapper with ufoo bootstrap and bus identity. |
uagy | Antigravity wrapper with ufoo bootstrap and bus identity. |
ucode | Native ufoo coding-agent CLI/TUI. |
Initialize a project and open the chat dashboard:
cd your-project
ufoo init --targets context,bus
ufoo
Launch agents from chat:
> /launch codex reviewer
> /launch claude builder
> /launch ucode fixer
> @reviewer inspect the current diff and list release risks
Or launch wrappers directly inside a project:
uclaude
ucodex
uagy
ucode
Use global chat mode to switch between registered projects:
ufoo -g
For MCP-capable clients, configure the global stdio bridge once:
ufoo mcp
The MCP bridge connects to the home-scoped global controller daemon and routes project-scoped tools through the global project registry. It is not a separate per-project MCP server mode.
The normal workflow is to enter chat first, then launch agents and run project commands inside the dashboard:
ufoo
ufoo -g
ufoo opens the current project chat. ufoo -g opens global chat for switching between registered projects. The project daemon is started as needed.
Project configuration is stored in .ufoo/config.json. ucode provider credentials are stored globally in ~/.ufoo/config.json and merged at load time.
Common project settings:
{
"launchMode": "auto",
"agentProvider": "codex-cli",
"controllerMode": "main",
"codexInternalThreadMode": "api",
"codexAuthPath": "",
"codexOauthRefreshWindowSec": 300,
"claudeOauthProfile": "",
"claudeOauthTokenPath": "",
"claudeOauthRefreshWindowSec": 300,
"routerProvider": "",
"routerModel": "",
"agentModel": "",
"autoResume": true
}
Supported launchMode values: auto, internal, tmux, terminal, and host. controllerMode accepts main, shadow, loop, and legacy compatibility values.
Global ucode settings:
{
"ucodeProvider": "openai",
"ucodeModel": "gpt-4.1",
"ucodeBaseUrl": "https://api.openai.com/v1",
"ucodeApiKey": "sk-...",
"ucodeAgentDir": ""
}
Run a local entry directly if the linked binary is not on PATH:
node bin/ufoo.js --help
If Codex cannot write under its default home, point it at a project-local directory before launching chat or agents:
export CODEX_HOME="$PWD/.ufoo/codex"
ufoo
For Codex-friendly notifications, prefer bus helpers over raw terminal text injection:
ufoo bus alert codex:abc123 --daemon
ufoo bus listen codex:abc123
该项目提供了一个开源的AI工作流管理和协调系统,支持多代理工作空间协议,事件总线和协调功能,帮助开发者高效管理项目和协调团队。然而,项目的star数较少,可能需要更多的维护和更新。
该工具使用 NOASSERTION 协议,商用场景请仔细阅读协议条款,必要时咨询法律意见。
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
📄 NOASSERTION — 请查阅原始协议条款了解具体使用限制。
总体来看,开源AI工作流 是一款质量良好的Agent工作流,在同类工具中具备一定竞争力。AI Skill Hub 将持续追踪其更新动态,建议收藏备用,结合自身场景选择合适时机引入使用。
| 原始名称 | ufoo |
| 原始描述 | 开源AI工作流:Multi-agent workspace protocol - context sharing, event bus, and coordination fo。⭐15 · JavaScript |
| Topics | workflowjavascript |
| GitHub | https://github.com/Icyoung/ufoo |
| License | NOASSERTION |
| 语言 | JavaScript |
收录时间:2026-05-24 · 更新时间:2026-05-30 · License:NOASSERTION · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端