agent-board
Open-source multi-agent task board for OpenClaw. Kanban + DAG dependencies + MCP server + auto-retry + audit trail. Built for autonomous AI agent teams.
claude mcp add --transport stdio quentintou-agent-board node dist/index.js \ --env TEMPLATES_DIR="Custom templates directory (default: ./templates)" \ --env OPENCLAW_HOOK_URL="OpenClaw webhook URL for agent notifications" \ --env AGENTBOARD_API_KEYS="Comma-separated key:agentId pairs for API authentication (optional)" \ --env OPENCLAW_HOOK_TOKEN="Bearer token for OpenClaw webhook calls (optional)" \ --env AGENTBOARD_WEBHOOK_SECRET="Secret for HMAC-SHA256 webhook signing (optional)"
How to use
Agent Board is a MCP-enabled multi-agent task orchestration server designed to manage work across AI agents in a Kanban-style board. It exposes a REST API and a set of MCP tools that agents can use to interact with tasks, comments, and workflow state transitions. The server enforces DAG dependencies, auto-retries failed tasks, supports task chaining, and maintains a complete audit trail. It also integrates with OpenClaw for triggering agent wakeups and supports HMAC-SHA256 signing for outbound webhooks, ensuring secure inter-agent communications. To get started, run the server and access the dashboard or REST API to create tasks, define templates, and configure webhooks and authentication as needed. The MCP tooling allows agents to perform actions such as reading task threads, posting comments, and retrieving full task context through dedicated MCP tools, enabling seamless agent-to-agent collaboration without bespoke orchestration code.
How to install
Prerequisites:
- Git
- Node.js (14+ or compatible) and npm
- Optional: Docker if you prefer containerized deployment
Install steps:
-
Clone the repository git clone https://github.com/quentintou/agent-board.git cd agent-board
-
Install dependencies npm install
-
Build the project (if a build step is required by the project) npm run build
-
Run the MCP server npm start
-
Access the server
- Dashboard: http://localhost:3456
- REST API: http://localhost:3456/api
Environment configuration (optional but common):
- AGENTBOARD_API_KEYS: Comma-separated key:agentId pairs for API authentication (e.g., sk-abc123:agent1,sk-def456:agent2)
- OPENCLAW_HOOK_URL: OpenClaw webhook URL for agent notifications (default: http://localhost:18789/hooks/agent)
- OPENCLAW_HOOK_TOKEN: Bearer token for OpenClaw webhook calls
- AGENTBOARD_WEBHOOK_SECRET: Secret for HMAC-SHA256 webhook signing (enables signature headers)
- TEMPLATES_DIR: Custom templates directory (default: ./templates)
Additional notes
Tips and common considerations:
- If you enable AGENTBOARD_WEBHOOK_SECRET, outbound webhooks will include X-AgentBoard-Signature headers for verification by recipients.
- Ensure OPENCLAW_HOOK_URL is accessible from the Agent Board host when integrating with OpenClaw.
- Use AGENTBOARD_API_KEYS to restrict access; when omitted, authentication may be disabled per backward compatibility.
- The MCP tooling (e.g., board_list_comments, board_add_comment, board_get_task_thread) enables MCP-based interactions without HTTP calls.
- Review the default ports and data directories (e.g., port 3456 and data directory) and customize via command-line flags or environment configuration as needed.
Related MCP Servers
mcp
🤖 Taskade MCP · Official MCP server and OpenAPI to MCP codegen. Build AI agent tools from any OpenAPI API and connect to Claude, Cursor, and more.
miro
MCP server for controlling Miro whiteboards with AI assistants
storybook
MCP server for Storybook - provides AI assistants access to components, stories, properties and screenshots. Built with TypeScript and Model Context Protocol SDK.
forensic-log
High-performance MCP server for log analysis. Give Claude the ability to analyze massive log files with SIMD-accelerated parsing. 5-50x faster than awk for aggregations.
aguara
MCP server for Aguara. Gives AI agents security scanning as a tool — checks skills, plugins, and configs before install.
claude-critical-rules
MCP server for Claude AI providing automatic enforcement of critical rules - Prevents 96 documented failure patterns