Your agent just spent $47 while you slept.
Podwatch catches runaway costs and risky behavior before they hit your bill.
Free forever. No credit card. 30-second setup.
Agents burn money silently
$47 overnight. $135/week in context bleed. You find out when the invoice arrives.
Agents access everything
Your .env, your SSH keys, your credentials. One bad tool call away from exfiltration.
Agent memory degrades
Contradictions, bloat, drift. Your agent gets dumber every week and you can't see it.
How Podwatch protects you
Three layers of protection between your agent and your bank account.
Cost Guardrails
Set a daily budget. Get warned at 80%. Hard-stop at 100%. Your agent stops before your bill doesn't.
Security Monitoring
Every tool call classified. Exfiltration patterns detected. Suspicious sequences flagged in real time.
Config & Memory Health
8 automated health checks. Memory drift detection. AI-powered analysis of your agent's brain.
Protected in 30 seconds
No complex setup. No credit card. One command.
1. Install in 30 seconds
One command. No dependencies. No configuration files.
2. Your agent runs normally
Podwatch monitors every tool call, cost, and behavior pattern — invisibly.
3. Get protected
Budget alerts, hard stops, and security flags — before damage happens.
“I built Podwatch after my agent burned $200 in a single weekend without me noticing.”
Monitoring 27,000+ events across 50+ agents
Frequently Asked Questions
Set a daily spending limit for your agent. Podwatch tracks every API call cost in real time. At 80% of your budget, you get an alert. At 100%, the agent is hard-stopped — no more API calls until you approve. You stay in control, even when you're asleep.
Podwatch classifies every tool call your agent makes. It detects patterns like reading .env files followed by network requests (potential exfiltration), sudo commands, credential access, and suspicious shell sequences. Each event is tagged as safe, warning, or danger in real time.
Yes. The Podwatch plugin runs locally on your machine and collects data on-disk. Secrets are redacted before any data leaves your machine. The collector plugin is source-available — you can inspect every line of code. Your agent's internal logs never leave your control.
About 30 seconds. Sign up (free, no credit card), grab your API key, and run one command: npx podwatch YOUR_KEY. Your agent starts being monitored immediately.
The Free plan includes 2 instances (API keys), 7-day data retention, unlimited events, real-time cost tracking, health scores, and basic monitoring. Upgrade to Pro ($12/mo) for budget enforcement, security scanning, AI-powered analysis, and 90-day retention.
No. Podwatch hooks into your agent's lifecycle events asynchronously. It monitors tool calls, costs, and behavior without adding latency to your agent's operations. The plugin is lightweight and designed to be invisible to your agent's performance.
Your agent is running right now.
Do you know what it's doing?
Every minute without guardrails is a minute your agent could be burning money, leaking secrets, or drifting off course.
Start Protecting — FreeFree forever. No credit card. 30-second setup.