agent-board

Overview

Agent Board is a task orchestration system designed for AI agent teams, providing a Kanban board interface with drag-and-drop functionality. It supports DAG task dependencies, automatic retry mechanisms for failed tasks, and built-in audit trails. The system exposes 12 MCP tools allowing AI agents to interact with the board without direct HTTP calls. It integrates specifically with OpenClaw multi-agent setups but can work with any LLM-based agent team through REST API or MCP protocol.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• board_list_projects

  List all projects or filter by status and owner

  Try:Show me all active projectsTry:List projects owned by alice
• board_get_project

  Get a specific project and its tasks

  Try:Get details for project 123Try:Show me the website redesign project
• board_create_project

  Create a new project with name, owner, and description

  Try:Create a new project called Mobile AppTry:Set up a project called Website Redesign for the agency
• board_list_tasks

  List tasks with optional filtering by project, assignee, status, or tag

  Try:Show me all tasks assigned to content-creatorTry:List all high priority tasks in the backlog
• board_get_task

  Get details of a specific task

  Try:Get details for task 456Try:Show me the research task findings
• board_create_task

  Create a new task with dependencies, review gates, and task chaining

  Try:Create a task to write landing page copy that depends on researchTry:Set up a task that needs review before completion
• board_update_task

  Update task fields like title, assignee, priority, or tags

  Try:Change the assignee of task 123 to design-agentTry:Update the priority of task 456 to high
• board_move_task

  Move a task to a different column, enforcing dependencies and quality gates

  Try:Move task 123 to the doing columnTry:Submit task 456 for review
• board_add_comment

  Add a comment to a task, which triggers webhook notifications

  Try:Add a comment to task 123 with the research findingsTry:Post an update on task 456 about progress
• board_get_task_thread

  Get a task's context and all comments in its thread

  Try:Show me the full thread for task 123Try:Get all comments on the design task
• board_get_dependencies

  List dependencies and blockers for a task

  Try:Show me what tasks are blocking task 123Try:List all dependencies for the research task
• board_get_dependents

  List tasks depending on a specific task

  Try:Show me all tasks that depend on task 123Try:What tasks are waiting for the research to complete

Comparable tools

Installation

git clone https://github.com/quentintou/agent-board.git
cd agent-board
npm install
npm run build
npm start

{
  "mcpServers": {
    "agent-board": {
      "command": "node",
      "args": ["dist/index.js"],
      "env": {
        "AGENTBOARD_API_KEYS": "sk-abc123:agent1,sk-def456:agent2"
      }
    }
  }
}

FAQ

Can Agent Board work with non-OpenClaw agents?

Yes, Agent Board can work with any LLM-based agent through REST API or MCP tools. It's designed as a general orchestration layer for AI agent teams.

How are task dependencies enforced?

Tasks cannot be moved to 'doing' until all their dependencies are marked as 'done'. The system includes cycle detection to prevent deadlocks in dependency chains.

Compare agent-board with