openclaw

OpenClaw Integration Guide

OpenClaw connects to external tools and environments using the Model Context Protocol (MCP) over stdio.

---

1. Installation

Install OpenBoard globally via npm:

npm install -g openboard

---

2. OpenBoard Startup

Start the local OpenBoard web dashboard and server:

openboard start

For AI agent operations via MCP, OpenClaw runs openboard mcp directly over stdio.

---

3. MCP Configuration

Configuration File (~/.openclaw/config.json)

Add OpenBoard under mcpServers in your OpenClaw configuration file:

{
  "mcpServers": {
    "openboard": {
      "command": "openboard",
      "args": ["mcp"]
    }
  }
}

---

4. Verification

Apply changes and verify tool registration:

openclaw gateway restart
openclaw mcp list

Verify that openboard is active and all 13 tools are available.

---

5. Example Workflows & Prompts

Multi-Step Deployment Topology Workflow

User:
List my OpenBoard whiteboards.

OpenClaw:
[Calls list_boards]
Found 2 boards: "Sprint Planning", "Deployment Architecture".

User:
Inspect the canvas on "Deployment Architecture".

OpenClaw:
[Calls get_canvas_state and get_canvas_screenshot on the target board]
Retrieved 5 shapes. Kubernetes cluster is currently disconnected from Redis.

User:
Connect the Kubernetes cluster to Redis with an arrow labeled "Cache Reads/Writes".

OpenClaw:
[Calls create_shapes with arrow binding from k8s_cluster to redis_cache]
Created arrow connecting Kubernetes to Redis with label "Cache Reads/Writes".

---

6. Troubleshooting

• Restart gateway: Always run openclaw gateway restart after editing config.json.
• Binary discovery: Ensure openboard is installed globally (npm install -g openboard).
• Clean stdio: OpenBoard ensures stdout contains only JSON-RPC protocol frames.