Open Source

Your agent crashed at 3 AM. How long until you noticed?

AgentWatch monitors your autonomous AI agents 24/7. Heartbeat checks, cost tracking, smart alerting, and auto-restart. Open source collector. Install in 60 seconds.

Free for up to 3 agents. No credit card required.

Sound familiar?

My research agent entered a retry loop and burned $47 in API calls before I woke up.

Cost runaway

My Telegram bot was down for 6 hours. Nobody told me — I found out from a user.

Silent failure

I'm running 5 agents on a Mac Mini. I have no idea if they're all still working.

Fleet blindness

Your Agent Fleet at a Glance

Real-time status, cost tracking, and uptime for every agent.

Fleet Overview
2 healthy1 degraded1 down
healthycode-reviewer
Uptime: 99.8%Cost: $1.23/dayLast seen: 12s ago
healthyemail-organizer
Uptime: 99.5%Cost: $0.87/dayLast seen: 8s ago
degradedtelegram-bot
Uptime: 97.2%Cost: $0.34/dayLast seen: 45s ago
downresearch-agent
Uptime: 91.1%Cost: $2.15/dayLast seen: 14m ago

Not Datadog. Not LangSmith.

Existing tools weren't built for autonomous AI agents.

Built for servers

Datadog / Grafana

Monitors CPU, memory, containers. Doesn't understand what an "agent" is. $23+/host/month.

Built for traces

LangSmith / Langfuse

Traces individual LLM calls. Doesn't tell you if your agent is alive or what it cost today.

Built for agents

AgentWatch

Is my agent alive? What did it cost? What happened while I slept? Get alerted when things break.

Everything you need to monitor AI agents

Purpose-built for autonomous agents. Not a generic infrastructure tool.

💓
Heartbeat Monitoring

Know within 60 seconds when an agent goes down. Green/yellow/red status for every service.

💰
Cost Tracking

See exactly what each agent costs per hour, day, and month. Get alerted before a runaway agent drains your budget.

🔄
Auto-Restart

Crashed agent? Collector restarts it automatically and tells you what happened.

🔔
Smart Alerting

Telegram, Slack, email, SMS. Escalation chains so you never miss a critical failure.

📋
Event Timeline

See everything your agent did while you were away. Searchable, filterable, color-coded.

🖥️
Multi-Agent Fleet

One dashboard for all your agents across all your machines. Fleet-wide health at a glance.

How it works

1

Install the collector

One command. Auto-discovers your running agents. No config files, no Docker, no YAML.

npx @agentwatch/collector init
2

Agents get monitored

The collector runs alongside your agents — heartbeat checks every 60s, cost tracking, process monitoring. Lightweight: <10MB, <30MB RAM.

3

Get alerted, not surprised

Agent down? Cost spike? Silent crash? Get a Telegram message, Slack alert, or SMS before your users notice.

Simple pricing

The collector is free and open source. Forever.

Free

$0

3 agents, 3-day retention, dashboard access

Pro

$29/mo

10 agents, 30-day retention, all alert channels

Team

$79/mo

25 agents, 90-day retention, team seats, RBAC

Open source. Fully inspectable.

The collector runs on your machine, next to your agents. You should be able to read every line of code and inspect every byte that leaves your machine. That's why it's MIT-licensed.

View on GitHub

Stop guessing. Start watching.

Join the waitlist for early access. Be the first to know when AgentWatch launches.