The AI Policy Layer
for Agentic Ecosystems
Limits intercepts every agentic action at the edge, centralizes and versions stateful policies, and replaces prompt-driven constraints with deterministic rules
## Refund Policy Instructions
When a customer requests a refund via
Stripe, you MUST follow these rules:
1. NEVER process refunds over $500
without explicit manager approval
2. ALWAYS verify the customer identity
before processing any refund
3. Do NOT process more than 10 refunds
per hour to prevent abuse
4. If the refund amount seems suspicious,
stop and ask for human review
5. Log every refund attempt for audit
# How do we actually enforce this?
# The agent just ignores it sometimes...import { Limits } from '@limits/sdk';
// Check policy before processing refund
const decision = await limits.check('stripe.refund', {
amount: 500,
customerId: 'cus_123'
});
if (decision.allowed) {
await stripe.refunds.create({ amount: 500 });
} else if (decision.requiresApproval) {
await limits.requestApproval(decision.id);
}Control agents the right way
Stop writing fragile prompts. Define policies that work across all your agents, tools, and workflows.
Unbreakable Rules
Prompts are just suggestions. Code is law. Enforce hard limits—like 'Refunds must be under $500' — that your AI cannot ignore, trick, or bypass.
Drop-in Safety Layer
Secure your agents without rewriting them. Limits sits in the middle, checking every single action instantly. You get enterprise-grade security without changing your existing code.
Human Approvals
Some actions can't be undone. Force your agent to ask for permission before touching live payments or production code, turning potential disasters into simple approval tickets.
Usage Limits & Quotas
Set rate limits, daily quotas, and usage caps per agent, per tool, or per user. Prevent runaway costs and resource exhaustion.
Simple, usage-based pricing
Start free, scale as you grow.
Hacker
For indie hackers and side projects
- Unlimited policies
- Unlimited agents
- Up to 1,000 policy checks/mo
- 1 Seat
- 3 days logs
- Email support
Team
For teams building AI products
- Unlimited policies
- Unlimited agents
- Up to 100,000 policy checks/mo
- Up to 5 Seats
- 30 days logs
- Human approval workflows
- Slack Notifications
- Audit Logs
- Priority Support
Enterprise
For organizations at scale
- Everything in Team, plus:
- Unlimited policy checks
- Unlimited seats
- Custom policy engine
- SSO & RBAC
- Dedicated support
- SLA guarantee
- On-premise option
Frequently Asked Questions
Everything you need to know about Limits.
Use our AI Copilot to describe your policies in plain English and have them generated automatically. Or use our visual Policy Builder UI to create rules with drag-and-drop.
Policy checks typically add ~50ms of latency. For most applications.
Yes! Human-in-the-loop is a core feature. You can define policies that pause execution and request approval via Slack, email, or our dashboard before sensitive operations proceed.
Take control of your agents
No credit card required · 1,000 free policy checks/month