Skip to main content

​
Agent Testing Example - Personal AI Assistant

This directory gives an example of running Arklex User Simulator CLI with a personal AI assistant agent for the Moltbot/OpenClaw use case. You can follow the example to evaluate your own OpenClaw agent.
We recommend using a sandbox environment for this agent, as real actions could take place. Please carefully grant access to tools and APIs, and ensure you understand the permissions being provided.

​
Overview

Moltbot (OpenClaw) is an open-source personal AI assistant that helps users manage daily tasks, smart home devices, messaging, calendar, and files. This example demonstrates how to simulate and evaluate conversational interactions with a personal assistant agent.

​
Data Preparation

This example includes a data folder with knowledge files covering:
  • Reminders and task management
  • Smart home device control
  • Cross-platform messaging (WhatsApp, Telegram, Slack, Discord)
  • Calendar and scheduling
  • File and document management
  • Memory and user preferences
  • Quick lookups and information retrieval
You can add, remove, or update knowledge files (.md, .pdf, .doc, etc.) in the data/ folder to match your use case.

​
OpenClaw Instance Setup

Use this for production evaluation against your actual OpenClaw deployment.

​
Prerequisites

  1. Install and configure OpenClaw: https://openclaw.ai
  2. Enable the HTTP Chat Completions endpoint in ~/.openclaw/openclaw.json: 
    {
  "gateway": {
    "http": {
      "endpoints": {
        "chatCompletions": {
          "enabled": true
        }
      }
    }
  }
}

​
Steps

1

Start the OpenClaw gateway

openclaw gateway --port 18789 --verbose
2

Set environment variables

export OPENAI_API_KEY="<YOUR_OPENAI_API_KEY>"
export OPENCLAW_TOKEN="<YOUR_OPENCLAW_GATEWAY_TOKEN>"
Find your token in ~/.openclaw/openclaw.json under gateway.auth.token.
3

Use the local config

Use the local config agent_config.json.
4

Run the simulator

./run_arksim.sh run ./examples/moltbot/config.yaml

​
Troubleshooting

Verifying OpenClaw Connection Test the endpoint before running the simulator: 
curl -sS "http://127.0.0.1:18789/v1/chat/completions" \
  -H "Authorization: Bearer $OPENCLAW_TOKEN" \
  -H "Content-Type: application/json" \
  -H "x-openclaw-agent-id: main" \
  -d '{"model":"openclaw","messages":[{"role":"user","content":"Hello"}]}'