openclaw-mission-control

SKILL.md

Mission Control

Coordinate a team of AI agents using a Kanban-style task board with HTTP API.

Overview

Mission Control lets you run multiple AI agents that collaborate on tasks:

Team Lead: Creates and assigns tasks, reviews completed work

Worker Agents: Poll for tasks via heartbeat, execute work, log progress

Kanban Board: Visual task management at http://localhost:8080

HTTP API: Agents interact via REST endpoints

Local Storage: All data stored in JSON files — no external database needed

Quick Start

1. Install the Kanban Board
# Clone the Mission Control app
git clone https://github.com/0xindiebruh/openclaw-mission-control.git
cd mission-control

# Install dependencies
npm install

# Start the server
npm run dev

The board runs at http://localhost:8080.

2. Configure Your Agents

Edit lib/config.ts to define your agent team:

export const AGENT_CONFIG = {
  brand: {
    name: "Mission Control",
    subtitle: "AI Agent Command Center",
  },
  agents: [
    {
      id: "lead",
      name: "Lead",
      emoji: "🎯",
      role: "Team Lead",
      focus: "Strategy, task assignment",
    },
    {
      id: "writer",
      name: "Writer",
      emoji: "✍️",
      role: "Content",
      focus: "Blog posts, documentation",
    },
    {
      id: "growth",
      name: "Growth",
      emoji: "🚀",
      role: "Marketing",
      focus: "SEO, campaigns",
    },
    {
      id: "dev",
      name: "Dev",
      emoji: "💻",
      role: "Engineering",
      focus: "Features, bugs, code",
    },
    {
      id: "ux",
      name: "UX",
      emoji: "🎨",
      role: "Product",
      focus: "Design, activation",
    },
    {
      id: "data",
      name: "Data",
      emoji: "📊",
      role: "Analytics",
      focus: "Metrics, reporting",
    },
  ] as const,
};

3. Seed the Database (First Run)

Initialize the agents in the database:

curl -X POST http://localhost:8080/api/seed

This creates agent records from your lib/config.ts configuration. Safe to run multiple times — it only adds missing agents.

4. Configure OpenClaw Multi-Agent Mode

Add each agent to your ~/.openclaw/config.json:

{
  "sessions": {
    "list": [
      {
        "id": "main",
        "default": true,
        "name": "Lead",
        "workspace": "~/.openclaw/workspace"
      },
      {
        "id": "writer",
        "name": "Writer",
        "workspace": "~/.openclaw/workspace-writer",
        "agentDir": "~/.openclaw/agents/writer/agent",
        "heartbeat": {
          "every": "15m"
        }
      },
      {
        "id": "growth",
        "name": "Growth",
        "workspace": "~/.openclaw/workspace-growth",
        "agentDir": "~/.openclaw/agents/growth/agent",
        "heartbeat": {
          "every": "15m"
        }
      },
      {
        "id": "dev",
        "name": "Dev",
        "workspace": "~/.openclaw/workspace-dev",
        "agentDir": "~/.openclaw/agents/dev/agent",
        "heartbeat": {
          "every": "15m"
        }
      }
    ]
  }
}

Key fields:

id: Unique agent identifier (must match an agent ID in lib/config.ts)

workspace: Agent's working directory for files

agentDir: Contains SOUL.md, HEARTBEAT.md, and agent personality

heartbeat.every: Polling frequency (e.g., 5m, 15m, 1h)

5. Set up Agent Heartbeats

Each worker agent needs a HEARTBEAT.md in their agentDir:

# Agent Heartbeat

## Step 1: Check for Tasks

```bash
curl "http://localhost:8080/api/tasks/mine?agent=writer"
```

Step 2: Pick up todo tasks

curl -X POST "http://localhost:8080/api/tasks/{TASK_ID}/pick" \
  -H "Content-Type: application/json" \
  -d '{"agent": "writer"}'

Step 3: Log Progress

curl -X POST "http://localhost:8080/api/tasks/{TASK_ID}/log" \
  -H "Content-Type: application/json" \
  -d '{"agent": "writer", "action": "progress", "note": "Working on..."}'

Step 4: Complete Tasks

curl -X POST "http://localhost:8080/api/tasks/{TASK_ID}/complete" \
  -H "Content-Type: application/json" \
  -d '{
    "agent": "writer",
    "note": "Completed! Summary...",
    "deliverables": ["path/to/output.md"]
  }'

Step 5: Check for @Mentions

curl "http://localhost:8080/api/mentions?agent=writer"

Mark as read when done.

Create the agent directories:

```bash
mkdir -p ~/.openclaw/agents/{writer,growth,dev,ux,data}/agent
mkdir -p ~/.openclaw/workspace-{writer,growth,dev,ux,data}

Task Lifecycle

backlog → todo → in_progress → review → done
   │        │         │           │
   │        │         │           └─ Team Lead approves
   │        │         └─ Agent completes (→ review)
   │        └─ Agent picks up (→ in_progress)
   └─ Team Lead prioritizes (→ todo)

Team Lead Operations

Creating a Task

curl -X POST http://localhost:8080/api/tasks \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Task title",
    "description": "Detailed description",
    "priority": "high",
    "assignee": "writer",
    "tags": ["tag1", "tag2"],
    "createdBy": "lead"
  }'

Priority: urgent, high, medium, low

Moving to Todo

curl -X PATCH "http://localhost:8080/api/tasks/{id}" \
  -H "Content-Type: application/json" \
  -d '{"status": "todo"}'

Approving Completed Work

curl -X PATCH "http://localhost:8080/api/tasks/{id}" \
  -H "Content-Type: application/json" \
  -d '{"status": "done"}'

Adding Deliverable Path

curl -X PATCH "http://localhost:8080/api/tasks/{id}" \
  -H "Content-Type: application/json" \
  -d '{"deliverable": "path/to/file.md"}'

Worker Agent Operations

Picking Up Tasks

curl -X POST "http://localhost:8080/api/tasks/{id}/pick" \
  -H "Content-Type: application/json" \
  -d '{"agent": "{AGENT_ID}"}'

Logging Progress

curl -X POST "http://localhost:8080/api/tasks/{id}/log" \
  -H "Content-Type: application/json" \
  -d '{
    "agent": "{AGENT_ID}",
    "action": "progress",
    "note": "Updated the widget component"
  }'

Actions: picked, progress, blocked, completed

Completing a Task

curl -X POST "http://localhost:8080/api/tasks/{id}/complete" \
  -H "Content-Type: application/json" \
  -d '{
    "agent": "{AGENT_ID}",
    "note": "Completed! Summary of changes...",
    "deliverables": ["docs/api.md", "src/feature.js"]
  }'

Deliverables render as markdown in the task view.

Comments & @Mentions

Adding a Comment

curl -X POST "http://localhost:8080/api/tasks/{id}/comments" \
  -H "Content-Type: application/json" \
  -d '{
    "author": "agent-id",
    "content": "Hey @other-agent, need your input here"
  }'

Checking for @Mentions

curl "http://localhost:8080/api/mentions?agent={AGENT_ID}"

Marking Mentions as Read

curl -X POST "http://localhost:8080/api/mentions/read" \
  -H "Content-Type: application/json" \
  -d '{"agent": "{AGENT_ID}", "all": true}'

API Reference

Tasks

Endpoint
Method
Description

/api/tasks
GET
List all tasks

/api/tasks
POST
Create new task

/api/tasks/{id}
GET
Get task detail

/api/tasks/{id}
PATCH
Update task fields

/api/tasks/{id}
DELETE
Delete task

/api/tasks/mine?agent={id}
GET
Agent's assigned tasks

/api/tasks/{id}/pick
POST
Agent picks up task

/api/tasks/{id}/log
POST
Log work action

/api/tasks/{id}/complete
POST
Complete task (→ review)

/api/tasks/{id}/comments
POST
Add comment

Agents & System

Endpoint
Method
Description

/api/agents
GET
List all agents

/api/seed
POST
Initialize agents (first run)

/api/mentions?agent={id}
GET
Get unread @mentions

/api/mentions/read
POST
Mark mentions as read

Files

Endpoint
Method
Description

/api/files/{path}
GET
Read deliverable content

Recommended Agent Team Structure

Agent
Role
Responsibilities

Lead
Team Lead
Strategy, task creation, approvals

Writer
Content
Blog posts, documentation, copy

Growth
Marketing
SEO, campaigns, outreach

Dev
Engineering
Features, bugs, code

UX
Product
Design, activation, user flows

Data
Analytics
Metrics, reports, insights

Configuration

Environment Variables

Create .env in your Mission Control app directory (optional):

PORT=8080

Data Storage

All data is stored locally in the data/ directory:

File
Contents

data/tasks.json
All tasks, comments, work logs

data/agents.json
Agent status and metadata

data/mentions.json
@mention notifications

Add data/ to your .gitignore — user data shouldn't be committed.

Example: Running a Multi-Agent Workflow

Lead creates task:

curl -X POST http://localhost:8080/api/tasks \
  -H "Content-Type: application/json" \
  -d '{"title": "Write Q1 Report", "assignee": "writer", "priority": "high"}'

Lead moves to todo:

curl -X PATCH http://localhost:8080/api/tasks/123 \
  -d '{"status": "todo"}'

Writer picks up via heartbeat:

curl -X POST http://localhost:8080/api/tasks/123/pick \
  -d '{"agent": "writer"}'

Writer completes:

curl -X POST http://localhost:8080/api/tasks/123/complete \
  -d '{"agent": "writer", "deliverables": ["reports/q1.md"]}'

Lead reviews and approves:

curl -X PATCH http://localhost:8080/api/tasks/123 \
  -d '{"status": "done"}'

Tips

Heartbeat frequency: 15 minutes is a good default

Priority order: Agents should work urgent → high → medium → low

Deliverables: Include all file paths modified in the task

@Mentions: Use to coordinate between agents on dependencies

Isolation: Each agent has its own workspace for safety

Storage: Data persists in data/ directory — back it up if needed

Resources

GitHub: https://github.com/0xindiebruh/openclaw-mission-control

Demo: See example agent setups in /examples