Fix AI agents continuously.
Or audit on demand.
Subscribe to the full fixing loop — variants, simulation, verification, ship — for $299/mo. Or pay $9 per audit, no commitment.
One audit on us. See what we find.
- 1 lifetime audit
- Full report — no gating
- Token-link sharing
- Try one optimization run
Run audits on demand. No subscription.
- $9 / audit, no commitment
- Bulk packs: 10 for $80, 50 for $400 (email us)
- Audit history retained, re-run any audit for $9
- Priority queue on cached vendors
The full loop. Audits, variants, simulation, verification.
- 15 audits / month
- 20 optimizations / month
- 1,000 conversations monitored
- Variant generation + simulation
- Regression suite + verification
- Scheduled cadence, multi-agent fleet
- Production deploy gates
- $9 / audit overage, $15 / opt overage
Fleet-scale loop. Custom probes. SSO, SLA, VPC.
- Volume-priced audits + optimizations
- Custom probe libraries
- SSO / SAML, audit logs
- Dedicated support + SLA
- On-prem / VPC option
Prices in USD. PAYG bills per audit. No long-term contracts on Pro.
Find what's broken.
35-probe behavioral battery against your production agent. Get a report with deductions, evidence, and a severity-ranked fix list. $9 per audit, no subscription, no SDK. Run one when you ship a change, after a customer complaint, or before a release.
Fix it. Verify it. Ship it.
Every audit becomes the input to the loop: variant generation, simulation against personas, regression suite, verification on production traffic, deploy gates. Pro $299/mo is for teams that want this running continuously, not just on demand.
Common questions.
Why pay-as-you-go instead of an audit subscription?
Audits are episodic. Some weeks you run zero, some weeks you run twenty after a regression hits. A monthly subscription bills you the same regardless. PAYG aligns price with usage and lets you stop and start without canceling anything. If you start running 5+ audits a month consistently, Pro at $299 (which includes 15 audits + the full fixing loop) becomes the better deal.
What counts as one audit?
One run of the 35-probe behavioral battery against one agent surface. Re-running the same agent (e.g., after you ship a fix) uses one audit. A multi-agent product is one audit per agent.
What counts as one optimization?
One full loop: diagnose → generate variants → simulate against personas → gate on metrics → decide. The diagnose step is itself an audit — a Pro plan running 20 optimizations also produces 20 audit-equivalent reports, plus 15 standalone audits on top.
Can I cancel Pro anytime?
Yes. No long-term contracts. Cancel and your access continues through the end of the billing period. Audit history and agents stay accessible.
What if I need more than 15 audits / month on Pro?
Audit overage at $9 / audit (parity with PAYG). A 30-audit month costs $299 + 15 × $9 = $434. If you're consistently running 50+ audits / month across a fleet, talk to us about Enterprise — that's a fleet pattern that prices to actual usage.
What's the difference between Eval and the loop?
Eval finds what's broken (audits + reports). The loop adds the fixing: variant generation against the failure modes the audit found, simulation testing against personas, verification on production traffic, and deploy gates. PAYG and Pro are different relationships with the same product family — one is on demand, the other is continuous.
Run your first audit, free.
10 minutes. No SDK. Decide whether to come back at $9 per audit or subscribe to the loop after you see the report.
Run free audit