OpenClaw Integration

Get Started

One skill. One server. One command. That’s it.

Install the Skill

This is the entire integration. One folder, two files - a markdown doc and a shell helper.

Tell OpenClaw: “Install the wireclaw skill from github.com/M64GitHub/WireClaw - it’s in the skill/wireclaw folder.”

Start a NATS Server

OpenClaw and WireClaw talk through NATS.

Tell OpenClaw: “Install the NATS server and CLI, then start the server.”

Then set the NATS server IP in WireClaw’s web config portal so it knows where to connect.

Want to run the NATS server on a different machine, or install manually? See detailed setup →

Talk to Hardware

Ask OpenClaw to use the WireClaw skill to discover devices on your network:

OpenClaw

Use the wireclaw skill to discover devices on my network

Found wireclaw-01 - ESP32-C6, v0.4.0, chip temp 33.2°C, 0 rules, 19 tools.

Set the LED to purple

Purple! 💜 30ms round trip. Can you see it?

That’s it. From chat to hardware in three steps. No SDK, no plugins, no config files. Just a skill.

How It Works

You just talk to OpenClaw. It handles the rest.

The skill teaches OpenClaw to talk to WireClaw. You just say what you want - OpenClaw translates it into hardware commands. Under the hood, they communicate through NATS - the same lightweight message bus that WireClaw already uses for its sensors and device mesh. Point both at a NATS server on any machine, and they find each other.

What You Can Build

Concrete ideas. Most take one or two commands.

CI Build Light

OpenClaw watches your GitHub Actions. Green LED on pass, red on fail. Push once - WireClaw handles the rest with a nats_value sensor and a persistent rule.

Morning Wake-Up Light

Warm orange LED at 7:30 AM, off at 8:00. One time-based rule on clock_hhmm. No cron, no cloud. The chip just knows.

Weather-Reactive Desk Lamp

OpenClaw checks the forecast, publishes temperature to NATS. WireClaw’s nats_value sensor picks it up - blue when cold, red when hot, green when perfect. Autonomous after setup.

Temperature Telegram Alert

NTC sensor on an ESP32 near your server rack. Rule fires a Telegram message when it crosses 40°C. No polling. No lambda. Just a rule on a chip.

Live Event Monitor

Subscribe to wireclaw-01.events and stream rule triggers, sensor readings, and status changes in real time. Debug rules, log data, or feed into dashboards - all through NATS.

Inside The Skill

One folder. Two files. That’s what teaches an AI agent to control hardware.

SKILL.md

Full tool reference - every parameter, every constraint, every edge case. Derived from WireClaw’s actual system prompt. OpenClaw generates correct tool calls on the first try.

SKILL.md - what OpenClaw sees

---
name: wireclaw
description: >
  Control WireClaw IoT devices on your network.
  Use when the user wants to interact with physical
  hardware: LEDs, GPIO pins, sensors, relays ...
tools:
  - Bash
  - Read
metadata:
  openclaw:
    requires:
      binaries:
        - nats
    env:
      - WIRECLAW_NATS_URL
---

# WireClaw - Physical World Automation for OpenClaw
# ... 19 tools, examples, constraints, patterns

It’s just markdown. OpenClaw reads it once, learns every tool, and generates correct NATS payloads. No plugins. No SDK. No config.

wc.sh

Convenience wrapper for NATS calls:

wc.sh - usage

wc.sh exec <device> <tool> [params]   # Execute a tool
wc.sh caps <device>                    # Query capabilities
wc.sh discover                         # Find all devices
wc.sh sub  <device>                    # Subscribe to events

Your AI Agent Gets a Body

First Contact