Stop runaway AI agent loops before they cost you.

LoopGuard is a runtime guardrail SDK for autonomous AI agents. Set token budgets, loop depth limits, and per-tool caps. Kill runaway sessions instantly from the dashboard.

$npm install @loopguard/sdk

Open dashboard View on npm

How it works

Install + wrap

npm install @loopguard/sdk, then wrap your LLM client or tool functions with createGuard(). Two lines of code.

Set limits

Configure token budgets, loop depth limits, and per-tool call caps. Limits are enforced locally — zero latency overhead.

Kill switch

When an agent trips a limit, LoopGuard throws and stops the loop. Flip the kill switch from the dashboard to stop a live agent instantly.

Pricing

Free

✓ Unlimited guarded actions

✓ Local policy enforcement

✓ Remote kill switch

✓ 1-day event retention

Get started

PRO

Pro

$49/mo

✓ Everything in Free

✓ 7-day event retention

✓ Replay logs (full forensics)

✓ Priority support

Upgrade to Pro