OpenClaw MCP Server Guide: Connect Any API in Under 10 Minutes

The Model Context Protocol (MCP) is the universal bridge between LLMs and your data. In OpenClaw, MCP servers are the primary way we give agents "hands" to interact with external APIs, databases, and local tools.

I've connected dozens of services to my personal fleet—from complex CRM APIs to simple local weather sensors. This guide shows you the fastest path to getting a new API connected and functional within your OpenClaw environment.

What is an MCP Server?

An MCP server is a small application that exposes "tools" to an MCP-compatible client (like OpenClaw). It handles the authentication and low-level API calls, presenting a standardized interface that the agent can understand and use.

Instead of writing custom logic for every agent, you write one MCP server (or use an existing one) and every agent in your fleet immediately knows how to use it.

Prerequisites

Before we start, ensure you have:

• OpenClaw installed and configured
• Node.js (v18+) or Python (3.10+) installed
• An API key for the service you want to connect

Step 1: Choose Your Server Type

You have three options for connecting an API:

1. Pre-built Servers: Check the MCP Server Directory first. Most common APIs (GitHub, Google Maps, Slack) already have servers.
2. MCPorter: Use OpenClaw's mcporter tool to bridge existing MCP servers.
3. Custom Implementation: Write a simple server in TypeScript or Python (takes ~5 minutes).

Step 2: Configuration via OpenClaw

The easiest way to add an MCP server to OpenClaw is through the openclaw.json configuration. OpenClaw supports both stdio and sse (HTTP) transports.

Example: Adding the Brave Search MCP Server

Edit your ~/.openclaw/openclaw.json (or the one in your workspace):

{
  "mcp": {
    "servers": {
      "brave-search": {
        "command": "npx",
        "args": ["-y", "@modelcontextprotocol/server-brave-search"],
        "env": {
          "BRAVE_API_KEY": "your_key_here"
        }
      }
    }
  }
}

Step 3: Verification

After updating the config, restart your OpenClaw gateway:

bash
openclaw gateway restart

Verify the server is active by asking your agent to list its tools:

bash
# In your chat interface
"What tools do you have available from the brave-search server?"

Step 4: Writing a Custom MCP Server (Fast Track)

If a server doesn't exist, use the Model Context Protocol SDK. Here's a minimal TypeScript implementation using the @modelcontextprotocol/sdk.

import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import { CallToolRequestSchema, ListToolsRequestSchema } from "@modelcontextprotocol/sdk/types.js";

const server = new Server({
  name: "my-custom-api",
  version: "1.0.0",
}, {
  capabilities: { tools: {} },
});

server.setRequestHandler(ListToolsRequestSchema, async () => ({
  tools: [{
    name: "fetch_data",
    description: "Fetch data from my internal API",
    inputSchema: {
      type: "object",
      properties: { id: { type: "string" } },
      required: ["id"],
    },
  }],
}));

server.setRequestHandler(CallToolRequestSchema, async (request) => {
  if (request.params.name === "fetch_data") {
    const { id } = request.params.arguments as { id: string };
    // Call your API here
    return { content: [{ type: "text", text: `Data for ${id}: ...` }] };
  }
  throw new Error("Tool not found");
});

const transport = new StdioServerTransport();
await server.connect(transport);

Advanced: Using mcporter for Ad-hoc Servers

OpenClaw includes mcporter, a skill designed to manage and call MCP servers dynamically without permanent configuration changes. This is perfect for testing.

bash
# Tell OpenClaw to use mcporter to call a remote server
openclaw exec "mcporter call --url http://localhost:3001/sse --tool get_balance --args '{\"user\": \"jkw\"}'"

Troubleshooting Common Issues

• Environment Variables: Ensure API keys are passed in the env block of openclaw.json, not just your shell.
• Pathing: Use absolute paths for command if npx or python isn't on the system path used by the daemon.
• Timeouts: For long-running API calls, adjust the timeoutMs in your server configuration.

FAQ