Skip to main content

​
Quickstart

Get ACME running in observational mode. Nothing changes on your system — you just get visibility.

​
Prerequisites

  • A running agent system (OpenClaw, LangChain, AutoGPT, or custom)
  • Shell access to the host machine
  • 5 minutes

​
Step 1: Install

curl -fsSL https://acmeagentsupply.com/install.sh | bash
This installs the ACME CLI tools. No daemons, no background processes yet.

​
Step 2: Run RadCheck

acme radcheck
RadCheck performs a read-only scan of your agent environment and produces a reliability score from 0-100.
RadCheck examines:
  • Stall risk — patterns that indicate potential hangs
  • Silence gaps — periods of unexpected inactivity
  • Compaction pressure — context window utilization
  • Hygiene — configuration best practices
Nothing is modified. It’s purely observational.

​
Step 3: Interpret Your Score

Score RangeStatusWhat It Means
80-100🟢 GREENHealthy. Low immediate risk.
60-79🟡 YELLOWWatch closely. Some risk factors present.
40-59🟠 ORANGEAttention needed. Reliability concerns.
0-39🔴 REDCritical. High risk of silent failure.

​
Step 4: Start Sentinel (Optional)

If you want continuous runtime protection: 
acme sentinel start
Sentinel runs in the background, watching for stalls and silent failures. It doesn’t take autonomous action — it alerts and provides context.

​
Next Steps

RadCheck Deep Dive

Understand exactly what each score component means.

Sentinel Overview

Learn how runtime protection works.

​
Need Help?

  • Live Chat: Click the widget in the bottom-right corner
  • Email: support@acmeagentsupply.com
  • GitHub Issues: For bugs and feature requests