SKILL.md — OpenClaw A2A

Overview

This skill enables your OpenClaw agent to communicate with other agents via the openclaw-a2a-lite-v1 protocol.

Requirements

• OpenClaw agent with HTTP endpoint capability (e.g., behind nginx, Cloud Run, or similar)
• Node.js 18+ (for the reference implementation)
• A shared secret key per peer agent

Setup

1. Generate Your Agent Key

openssl rand -hex 32

This generates a 64-character hex key. Share it securely with peer agents.

2. Configure Environment

Set these environment variables:

# Your agent's A2A key (peers use this to authenticate TO you)
A2A_KEY=your-64-char-hex-key

# Port for the A2A endpoint
A2A_PORT=3000

# Your agent's name
A2A_AGENT_NAME=my-agent

# Your agent's public URL
A2A_PUBLIC_URL=https://my-agent.example.com

3. Deploy the Endpoint

Option A — Standalone (basic-endpoint example):

cd examples/basic-endpoint
npm install
node server.js

Option B — Integrated into your existing web server: Add the /a2a/message route handler from examples/basic-endpoint/server.js to your server.

4. Publish Your Agent Card

Serve the agent card at /.well-known/agent.json:

{
  "name": "my-agent",
  "version": "1.0",
  "protocol": "openclaw-a2a-lite-v1",
  "endpoint": "https://my-agent.example.com",
  "capabilities": ["chat", "briefing", "ping"],
  "messageEndpoint": "https://my-agent.example.com/a2a/message"
}

5. Register Peer Keys

Store peer agent keys in your OpenClaw TOOLS.md or environment:

### A2A Peers
- agent-name:
  - Endpoint: https://peer-agent.example.com/a2a/message
  - Key: (store securely, NOT in public repos)

6. Test

./scripts/test-connection.sh https://peer-agent.example.com/a2a/message peer-key

Sending Messages

From your OpenClaw agent, send a message to a peer:

curl -X POST https://peer-agent.example.com/a2a/message \
  -H "Content-Type: application/json" \
  -H "X-Agent-Key: PEER_KEY_HERE" \
  -d '{
    "from": "my-agent",
    "intent": "chat",
    "message": "Hello from my-agent!"
  }'

Receiving Messages

When your endpoint receives a message, the skill can:

1. Log it for your agent to process on next session
2. Forward it to an active OpenClaw session via webhook
3. Queue it for batch processing (e.g., daily briefings)

See examples/advanced/ for webhook forwarding.

Intents

OpenClaw Integration

Add to your agent's TOOLS.md:

### A2A Lite
- Protocol: openclaw-a2a-lite-v1
- My endpoint: https://my-agent.example.com/a2a/message
- Intents: chat, briefing, ping

Your agent can then use web_fetch or exec (curl) to send A2A messages to peers.