什么是 Project Cerebro：多大脑控制界面？

Project Cerebro 是一个高度模块化的执行控制界面，专为需要监督多个专业智能体（称为“大脑”）的用户而设计。作为 Openclaw Skills 库的重要补充，它提供了一个集中的仪表盘和运行时，通过执行流（Execution Stream）管理智能体工作流。这个“单一真理来源”跟踪所有当前和过去的执行情况，允许手动干预和完全自主运行的结合。

其核心功能是让开发者能够为不同的智能体 ID 定义特定的范围和工具配置。无论您是使用默认的 Nexus 大脑进行通用编排，还是使用专业模块处理复杂的工作流，该工具都为智能体友好型的执行和管理提供了一个稳健的框架。

下载入口:https://github.com/openclaw/skills/tree/main/skills/zacharytgray/project-cerebro

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

2. 手动安装

将技能文件夹复制到以下位置之一

优先级：工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

Project Cerebro：多大脑控制界面 应用场景

• 通过单一统一的仪表盘管理多个专业 AI 智能体。
• 调度循环任务，自动将实例生成到执行流中。
• 监督智能体活动，可在手动运行触发器和每个大脑的自动模式之间切换。
• 通过集成的 OpenClaw 消息服务路由跨渠道通知。

1. 仪表盘使用提供的环境变量连接到 OpenClaw 网关。
2. 用户在配置文件中定义大脑，将特定的 OpenClaw 智能体 ID 映射到命名的范围。
3. 执行流作为所有单次和循环任务的中央账本。
4. 当手动或通过计划触发任务时，该技能会与网关通信以执行指令。
5. 状态更新和通知通过 OpenClaw 消息系统分发到 Discord 或 T@elegrimm 等渠道。

Project Cerebro：多大脑控制界面 配置指南

要开始使用 Openclaw Skills 集合中的这项技能，请按照以下步骤操作：

1. 克隆仓库并安装依赖：

npm install

2. 通过创建包含 OPENCLAW_GATEWAY_URL 和 OPENCLAW_TOKEN 的 .env 文件来配置您的环境。
3. 初始化您的配置文件：

cp config/brains.template.json config/brains.json
cp config/brain_targets.template.json config/brain_targets.json

4. （可选）填充示例任务：

npm run seed

Project Cerebro：多大脑控制界面 数据架构与分类体系

name: project-cerebro
version: 1.0.2
description: "Project Cerebro - Control Surface: modular multi-brain execution dashboard built on OpenClaw"
homepage: https://github.com/zacharytgray/project-cerebro
metadata: {"clawhub":{"category":"productivity","emoji":"??","requiresEnv":["OPENCLAW_GATEWAY_URL","OPENCLAW_TOKEN"],"exposesHttp":true,"fileUploads":true}}

Project Cerebro - Control Surface

A highly modular, agent-friendly execution control surface for running multiple OpenClaw “brains” (agents) with:

• an Execution Stream (source of truth)
• Recurring Tasks that spawn instances into the stream
• per-brain Auto Mode + manual Run
• channel-agnostic notifications via OpenClaw messaging

What this is

Cerebro is a dashboard + runtime that helps you supervise a set of specialized agents (“brains”).

Core model

• Execution Stream = source of truth for everything that will run / has run.
• One-time tasks execute when you click Run or when the brain is in Auto Mode.
• Recurring tasks spawn exactly one instance into the Execution Stream when due and execute regardless of Auto Mode.

Brains

• A brain is a named scope + tool/context configuration bound to an OpenClaw agent id.
• Nexus is the default brain: general-purpose orchestration and “everything else.”

Install / Setup (high level)

See SETUP.md for full instructions.

Required secrets / env vars

• OPENCLAW_GATEWAY_URL (e.g. ws://127.0.0.1:18789)
• OPENCLAW_TOKEN (OpenClaw Gateway token)

Quick outline:

1. Clone repo + install deps
2. Create .env from .env.example and set the required OpenClaw vars above
3. Copy templates:
   • config/brains.template.json → config/brains.json
   • config/brain_targets.template.json → config/brain_targets.json
4. (Optional) seed examples:
   • npm run seed

Modularity (agent-friendly)

This repo is designed for agents to customize quickly:

• Add/remove brains by editing config/brains.json and config/brain_targets.json.
• Optional Job module: opt in/out by adding/removing the Job brain + pages (see config/brains.jobs.addon.template.json).
• Recurring tasks are templates under config/seeds/.

Runtime behaviors (security transparency)

This package runs a local HTTP service (Fastify + React UI) and will:

• Expose a dashboard over HTTP on PORT (default 3000)
• Call the OpenClaw Gateway/CLI using the secrets you provide
• Send outbound messages using OpenClaw messaging to the per-brain {channel,target} destinations
• Accept file uploads via POST /api/upload and store them locally under ./data//intake/

Notes

• No secrets or personal IDs are committed; local config/DB are gitignored.
• Notifications are routed through OpenClaw’s channel system (Discord/T@elegrimm/Signal/etc.).
• Optional modules (Job tracker/profile) are opt-in; if you don’t include the Job brain, the UI hides job pages.