</> DEV Task Board

A real-time Kanban board designed for multi-agent AI workflows with OpenClaw. Assign tasks to AI agents, watch them work in real-time, and collaborate through persistent chat sessions.

✨ Features

🎯 Core Functionality

• Live Kanban Board — Real-time updates via WebSocket
• Multi-Agent Support — Assign tasks to different AI agents
• Auto-Spawn Sessions — Agents automatically activate when tasks move to "In Progress"
• Persistent Conversations — Back-and-forth chat with agents on each task
• Session Isolation — Each agent maintains separate context per task

🤖 AI Agents (Configurable via .env)

Icon	Agent	Focus
🤖	Main Agent	Coordinator, command bar chat (name configurable)
🏛️	Architect	System design, patterns, scalability
🔒	Security Auditor	SOC2, HIPAA, CIS compliance
📋	Code Reviewer	Code quality, best practices
🎨	UX Manager	User flows, UI consistency

💬 Communication

• Command Bar — Direct chat with your main agent from the header
• @Mentions — Tag agents into any task conversation
• Action Items — Questions, blockers, and completion tracking
• File Attachments — Paste images or attach documents

🔒 Security

• API key authentication for sensitive endpoints
• Secrets stored in environment variables
• CORS restricted to localhost
• Input validation and size limits
• Agent guardrails (filesystem boundaries, forbidden actions)

🚀 Quick Start

Prerequisites

• Docker & Docker Compose
• OpenClaw running locally

Installation

1. Clone the repository

   git clone https://github.com/yourusername/openclaw-taskboard.git
   cd openclaw-taskboard

2. Configure environment

   cp .env.example .env
   # Edit .env with your OpenClaw token and generate an API key

3. Start the task board

   docker-compose up -d

4. Open in browser

   http://localhost:8080
   

---

🤖 AI-Assisted Setup

The easiest way to set up the task board is to ask your OpenClaw agent to do it for you!

Connecting the Task Board (Channel Plugin)

Once the task board is running, prompt your OpenClaw agent:

I have the task board running at http://localhost:8080. 
Please onboard it as a channel plugin so you can receive 
messages from the command bar and spawn sub-agents when 
tasks move to "In Progress".

Your agent will:

1. Update the .env with the correct gateway URL and token
2. Verify the connection is working
3. Test the /tools/invoke API

Onboarding the Dev Team (Sub-Agents)

To set up the multi-agent dev team, prompt your agent:

I want to set up the dev team sub-agents (Architect, Security Auditor, 
Code Reviewer, UX Manager). Please configure them in OpenClaw so they 
can be spawned from the task board.

Your agent will guide you through:

1. Adding agent definitions to your OpenClaw config
2. Setting up the dev-team.md guardrails file
3. Configuring spawn permissions

---

👥 Setting Up the Dev Team

The task board works best with a team of specialized AI agents. Here's how to configure them:

Step 1: Configure Agents in OpenClaw

Add these agents to your OpenClaw config (~/.openclaw/openclaw.json):

{
  "agents": {
    "list": [
      {
        "id": "main",
        "name": "YourAgentName",
        "subagents": {
          "allowAgents": ["architect", "security-auditor", "code-reviewer", "ux-manager"]
        }
      },
      {
        "id": "architect",
        "name": "Architect",
        "identity": { "name": "Architect", "emoji": "🏛️" },
        "tools": { "profile": "coding", "deny": ["browser", "message"] }
      },
      {
        "id": "security-auditor",
        "name": "Security Auditor",
        "identity": { "name": "Security Auditor", "emoji": "🔒" },
        "tools": { "profile": "coding", "deny": ["browser", "message"] }
      },
      {
        "id": "code-reviewer",
        "name": "Code Reviewer",
        "identity": { "name": "Code Reviewer", "emoji": "📝" },
        "tools": { "profile": "coding", "deny": ["browser", "message"] }
      },
      {
        "id": "ux-manager",
        "name": "UX Manager",
        "identity": { "name": "UX Manager", "emoji": "🎨" },
        "tools": { "profile": "coding", "deny": ["message"] }
      }
    ]
  }
}

Step 2: Create Your Dev Team Guardrails

Copy the template to your OpenClaw workspace:

cp examples/dev-team-example.md ~/.openclaw/workspace/agents/dev-team.md

Edit dev-team.md to customize:

• Filesystem boundaries — Paths agents can access
• Compliance context — Your security requirements
• System prompts — Role-specific instructions

Step 3: Update Your Agent's TOOLS.md

Add this section to your main agent's TOOLS.md:

## Task Board Integration

**URL:** http://localhost:8080
**Container:** openclaw-taskboard

When spawning sub-agents from the task board:
1. Include the guardrails from `agents/dev-team.md`
2. Pass task context (title, description, recent comments)
3. Instruct agent to post updates as comments on the task card

### API Reference
- Create comment: `POST /api/tasks/{id}/comments`
- Move task: `POST /api/tasks/{id}/move`
- Create action item: `POST /api/tasks/{id}/action-items`

---

⚙️ Configuration

Environment Variables

Copy .env.example to .env and customize:

OpenClaw Integration

Variable	Description	Required
OPENCLAW_GATEWAY_URL	OpenClaw gateway URL	For AI features
OPENCLAW_TOKEN	OpenClaw API token	For AI features
TASKBOARD_API_KEY	API key for protected endpoints	Recommended

Project Configuration

These customize the agent guardrails and system prompts for your project:

Variable	Description	Example
PROJECT_NAME	Your project name	My SaaS App
COMPANY_NAME	Your company/team	Acme Corp
COMPANY_CONTEXT	Brief context for agents	fintech startup building payment APIs
ALLOWED_PATHS	Paths agents can access (comma-separated)	/home/user/myproject, /workspace
COMPLIANCE_FRAMEWORKS	Security/compliance context	SOC2, HIPAA, PCI-DSS

Branding

Variable	Description	Default
MAIN_AGENT_NAME	Your main agent's display name	Assistant
MAIN_AGENT_EMOJI	Emoji for main agent	🤖
HUMAN_NAME	Your display name	User
BOARD_TITLE	Page title	Task Board

Note: Without OpenClaw configured, the board works as a standard Kanban without AI agent automation.

OpenClaw Integration

📖 See OPENCLAW_SETUP.md for the full integration guide.

Quick overview:

1. Configure agents in OpenClaw (architect, security-auditor, code-reviewer, ux-manager)
2. Set your token in .env
3. Add task board handler to your agent's TOOLS.md

The task board will auto-spawn agent sessions when tasks move to "In Progress".

---

📋 Workflow

Backlog → In Progress → Review → Done
              ↓
           Blocked

1. Backlog — Tasks waiting to be started
2. In Progress — Agent session auto-spawns, work begins
3. Review — Agent completed work, awaiting approval
4. Done — Human approval required (cannot be set by agents)
5. Blocked — Waiting on external input

---

🧠 Session Isolation: One Agent, One Context

Each task card maintains its own isolated AI session. This is a game-changer for complex projects.

How It Works

Task #1: "Review Auth System"          Task #2: "Design API Schema"
         ↓                                      ↓
┌─────────────────────┐              ┌─────────────────────┐
│ Architect Session A │              │ Architect Session B │
│                     │              │                     │
│ • Knows about auth  │              │ • Knows about API   │
│ • Has auth context  │              │ • Has schema context│
│ • Separate memory   │              │ • Separate memory   │
└─────────────────────┘              └─────────────────────┘

Why This Matters

• No context bleed — Agent working on Task A won't confuse it with Task B
• Persistent conversations — Come back hours later, pick up where you left off
• True multitasking — Multiple agents can work on different tasks simultaneously
• Clean handoffs — Move task to Review, agent remembers everything when you ask follow-ups

---

👥 Multi-Agent Collaboration: @Mentions

Need a second opinion? Tag another agent into the conversation.

You: "@Security Auditor can you review the auth approach here?"
         ↓
Security Auditor receives context + responds in same thread

Scenario	Primary Agent	Tag In
Feature design needs security review	Architect	@Security Auditor
Code review found UX issues	Code Reviewer	@UX Manager
Complex decision needs multiple perspectives	Any	@Architect @Security Auditor

---

📋 Action Items

Action items track what needs attention with notification bubbles on cards:

Type	Trigger	Purpose
Question 🟡	Agent creates manually	Agent needs clarification
Completion 🟢	Auto on → Review	Work ready for approval
Blocker 🔴	Auto on → Blocked	Documents what's blocking

---

🏗️ Architecture

┌─────────────────────────────────────────────────────────┐
│                    Task Board UI                         │
│       WebSocket ←→ FastAPI Backend ←→ SQLite            │
└─────────────────────────┬───────────────────────────────┘
                          │ /tools/invoke
┌─────────────────────────┴───────────────────────────────┐
│                   OpenClaw Gateway                       │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐              │
│  │  Main    │  │ Architect│  │ Security │  ...         │
│  │  Agent   │  │          │  │ Auditor  │              │
│  └──────────┘  └──────────┘  └──────────┘              │
└─────────────────────────────────────────────────────────┘

---

🔌 API Endpoints

Tasks

• GET /api/tasks — List all tasks
• POST /api/tasks — Create task
• PATCH /api/tasks/{id} — Update task
• DELETE /api/tasks/{id} — Delete task
• POST /api/tasks/{id}/move — Move task to status

Comments

• GET /api/tasks/{id}/comments — Get comments
• POST /api/tasks/{id}/comments — Add comment

Action Items

• GET /api/tasks/{id}/action-items — Get action items
• POST /api/tasks/{id}/action-items — Create action item
• POST /api/action-items/{id}/resolve — Resolve item

Command Bar

• POST /api/jarvis/chat — Send message to main agent
• POST /api/jarvis/respond — Push response to command bar

WebSocket

• WS /ws — Real-time updates

---

🎨 Customization

Adding New Agents

Edit app.py:

AGENT_TO_OPENCLAW_ID = {
    "Your Agent": "your-agent-id",
    ...
}

AGENT_SYSTEM_PROMPTS = {
    "your-agent-id": "Your agent's system prompt...",
    ...
}

Update static/index.html for agent icon:

const AGENT_ICONS = {
    'Your Agent': '🚀',
    ...
};

---

📄 License

MIT License — see LICENSE

🙏 Credits

Built for the OpenClaw community.

---

Questions? Open an issue or check the OpenClaw Discord