Building Secure Agents with OpenClaw + XMTP

Overview

Build AI agents using OpenClaw's framework that can communicate via XMTP's quantum-resistant, privacy-preserving messaging protocol.

What is XMTP?

XMTP (Extensible Message Transport Protocol) is a quantum-resistant, privacy-preserving messaging protocol that enables secure communication between AI agents and users. It provides end-to-end encrypted messaging that protects against both current and future cryptographic threats while maintaining complete user privacy. XMTP allows agents to communicate directly with users without relying on centralized servers or infrastructure.

Why use XMTP with OpenClaw?

• Secure + Private by Default End-to-end encryption is built-in. Your agent's conversations are private between the agent and user - you don't need to implement encryption or worry about message security.
• Give Your Agent a Persistent Messaging Identity Your OpenClaw agent gets a persistent messaging identity that users can save, share, and return to - like a phone number or email, but decentralized.

Security Benefits of XMTP

Getting Started

Installation

XMTP and Convos are not in upstream OpenClaw main yet. Use the fork:

1. Fork xmtplabs/openclaw
2. Checkout branch feat/xmtp-and-convos-extensions (PR #2)
3. Build: pnpm install then pnpm build
4. Run OpenClaw from the repo (extensions are under extensions/ and are used automatically)

Do not use npm install -g openclaw or openclaw plugins install @openclaw/xmtp — they won’t have XMTP until it’s in main.

Step 1: Configure XMTP Channel

Set up the XMTP channel using OpenClaw's configuration wizard to enable secure messaging capabilities. This creates the foundation for your agent to send and receive encrypted messages.

# Run the configuration wizard
openclaw configure

When prompted:

• Choose environment: Production or Dev
• Choose keys:
  • Random - automatically generates a new messaging identity and encryption keys for your agent
  • Custom - enter your own existing keys if you already have them

The wizard will display your agent's public address. Share this address so others can message your agent via XMTP.

Manual Configuration (Optional)

Add to ~/.openclaw/openclaw.json:

{
  "channels": {
    "xmtp": {
      "enabled": true,
      "walletKey": "<hex-private-key>",
      "dbEncryptionKey": "<encryption-key>",
      "env": "production"
    }
  }
}

Step 2: Configure DM and Group Policies

Control how your agent handles incoming messages by setting policies for direct messages and group conversations. This determines who can interact with your agent and how it responds to different types of messages.

DM Policies:

{
  "channels": {
    "xmtp": {
      "dmPolicy": "pairing",  // Options: pairing, allowlist, open, disabled
      "allowFrom": ["0x123...", "0x456..."]  // Required when dmPolicy is "allowlist"
    }
  }
}

• pairing (default): Unknown senders get a pairing code; you approve
• allowlist: Only allow senders in allowFrom array
• open: Accept all incoming DMs
• disabled: Ignore all DMs

Group Policies:

{
  "channels": {
    "xmtp": {
      "groupPolicy": "open",  // Options: open, disabled, allowlist
      "groups": ["*"]  // Use "*" to allow all, or list specific conversation IDs
    }
  }
}

Step 3: Start Your Agent

Activate the agent to begin listening for incoming messages. Once started, your agent will automatically process and respond to any messages it receives through XMTP.

# Start the agent
openclaw start

# Or run in development mode with logs
openclaw start --debug

Your agent is now live! The console will show:

• ✅ Agent started
• 📬 Your agent's XMTP address
• 📨 Incoming messages
• 🤖 Agent responses

Step 4: Advanced Configuration

Customize additional XMTP settings to optimize your agent's behavior. These options control message handling, display settings, and security features.

{
  "channels": {
    "xmtp": {
      "enabled": true,
      "walletKey": "<hex-private-key>",
      "dbEncryptionKey": "<encryption-key>",
      "env": "production",
      "debug": false,
      "dmPolicy": "pairing",
      "groupPolicy": "open",
      "textChunkLimit": 4000,
      "name": "My Agent"
    }
  }
}

Configuration Options:

Multiple XMTP Identities:

{
  "channels": {
    "xmtp": {
      "accounts": {
        "main": {
          "walletKey": "<key1>",
          "dbEncryptionKey": "<key1>",
          "env": "production"
        },
        "support": {
          "walletKey": "<key2>",
          "dbEncryptionKey": "<key2>",
          "env": "production"
        }
      }
    }
  }
}

Complete Setup Example

Quick Start (Automated):

# Install OpenClaw and XMTP plugin
npm install -g openclaw
openclaw plugins install @openclaw/xmtp

# Configure XMTP (choose Random keys for quick setup)
openclaw configure

# Start your agent
openclaw start

Custom Setup:

# Install
npm install -g openclaw
openclaw plugins install @openclaw/xmtp

# Create config file at ~/.openclaw/openclaw.json
cat > ~/.openclaw/openclaw.json << EOF
{
  "channels": {
    "xmtp": {
      "enabled": true,
      "walletKey": "0x1234...",
      "dbEncryptionKey": "your-encryption-key",
      "env": "production",
      "dmPolicy": "pairing",
      "groupPolicy": "open",
      "name": "HelpBot"
    }
  }
}
EOF

# Start agent
openclaw start

Testing your agent

Use xmtp.chat (web)

1. Go to xmtp.chat
2. Connect with your wallet or create a new account
3. Start a new conversation
4. Enter your agent's XMTP address
5. Send a message to interact with your agent