RaunstenLABS
Home / Pricing
Pay for proof, not for promises.
Start free on your own projects. Upgrade when your AI feature has real users and a wrong answer starts costing real money.
For developers testing a prototype or a side project.
- Up to 25,000 scenario runs per month
- Functional and edge-case suites
- Fineness score and failure report
- CLI and one CI pipeline
- Community support
For product teams with an AI feature in production.
- Up to 500,000 scenario runs per month
- Full adversarial suite: jailbreaks, injection, leakage
- Release gates on unlimited pipelines
- Production monitoring on sampled traffic
- Slack alerts and email support
For companies running several AI products or agents.
- Up to 3,000,000 scenario runs per month
- Agent and tool-call safety suites
- Model-swap comparisons and cost tracking
- SSO, roles and audit log
- Shared Slack channel with our engineers
For regulated industries and large-scale deployments.
- Unlimited runs and custom attack libraries
- Deployment in your own cloud or VPC
- Audit-ready reports for risk and compliance teams
- Custom judges calibrated on your data
- Named assayer and uptime SLA
What’s inside each plan.
| Feature | Bench | Team | Scale | Enterprise |
|---|---|---|---|---|
| Scenario runs / month | 25K | 500K | 3M | Unlimited |
| Adversarial suite | — | Included | Included | Custom |
| Agent & tool-call testing | — | — | Included | Included |
| Production monitoring | — | Sampled | Sampled | Full traffic |
| CI release gates | 1 pipeline | Unlimited | Unlimited | Unlimited |
| Data retention | 14 days | 90 days | 1 year | Custom |
| Self-hosted deployment | — | — | — | Included |
| Support | Community | Shared Slack | Named assayer |
Questions we hear before every first assay.
What counts as a scenario run?
One scenario sent to your AI feature and graded. A suite of 10,000 scenarios run once counts as 10,000 runs. Production monitoring counts each sampled response as one run.
Do you pay for our LLM calls?
No. Scenarios are sent to your own endpoint, so model costs stay on your provider bill. We cover the cost of generating scenarios and running our judge models.
Do you train on our data?
Never. Your prompts, responses and documents are used only to run your assays. Enterprise customers can keep everything inside their own cloud.
What happens if we go over our limit?
Runs keep going and you’re billed for the overage at your plan’s rate. We’ll warn you at 80% so there are no surprises.
Can we cancel any time?
Yes. Monthly plans can be cancelled from your workspace settings, and you keep access until the end of the billing period.
Weigh your workload.
Drag the scale to your expected monthly volume. We’ll suggest a plan and send a quote that fits how you actually ship.
Request RSN-00000 received. Your quote is on its way.