Start with free tracing and AI safety analysis. Add development tools and production protection as you grow.
Start here
Understand your first application and agent
Included each month
Logging hard cap · no surprise overages
Development workflows
Build and evaluate production AI
Billed monthly
Included each month
Extra logging: $0.50 per 1,000 records
Production protection
Protect production and prepare evidence
Billed monthly
Included each month
Extra logging: $0.30 per 1,000 records
Custom agreements
Govern AI under an agreed operating contract
Included each month
Volumes and overage terms agreed with sales
Logging, AI classifications, and playground messages have separate allowances. One logged record is one inference event or one agent run; steps within an agent run do not add logging units. Logging overages do not increase the AI classification allowance. All amounts are in USD.
Core tracing and agent observability on every plan. More analysis and production tools when you need them.
| Feature | Starter | Growth | Pro | Enterprise |
|---|---|---|---|---|
| Monthly logged events + agent runs | 10K | 100K | 1M | Custom |
| Monthly AI classifications | 1K | 100K | 1M | Custom |
| Monthly playground messages | 25 | 500 | 2,500 | 10,000 |
| Extra logging per 1,000 records | Hard cap | $0.50 | $0.30 | Agreed |
| Default retention | 7 days | 30 days | 90 days | 365 days / custom |
| API keys | 2 | 10 | 50 | 1,000 |
| Ingestion rate limit | 60/min | 300/min | 1,000/min | 5,000/min |
| Built-in detector types | 3 | 5 | 7 | 8 |
| Core dashboard, Live Events, risk results, and audit logs | ✓ | ✓ | ✓ | ✓ |
| Distributed tracing and Model Analytics | ✓ | ✓ | ✓ | ✓ |
| Agent Overview and Runs | ✓ | ✓ | ✓ | ✓ |
| Harnesses and agent experiments | ✓ | ✓ | ✓ | ✓ |
| Agent feedback and issues | ✓ | ✓ | ✓ | ✓ |
| Drift alerts, incidents, and model comparison | — | ✓ | ✓ | ✓ |
| Prompts and versioning | — | ✓ | ✓ | ✓ |
| Evaluation datasets and runs | — | ✓ | ✓ | ✓ |
| Semantic caching | — | ✓ | ✓ | ✓ |
| Webhooks | — | 5 | 20 | 100 |
| PII analysis and factual accuracy detectors | — | — | ✓ | ✓ |
| Inline guardrails and rule management | — | — | ✓ | ✓ |
| Agent simulations | — | — | ✓ | ✓ |
| Executive dashboard | — | — | ✓ | ✓ |
| Benchmarks and compliance reports | — | — | ✓ | ✓ |
| Data exports (CSV, JSON, Excel, PDF, HTML) | — | — | ✓ | ✓ |
| Custom detectors | — | — | — | ✓ |
| SSO/SAML | — | — | — | By agreement |
| Availability commitment | — | — | 99.9% SLA | Agreed SLA |
| Support | Community | Priority | Dedicated |
Starter detectors: hallucination, policy violation, and confidence analysis. Growth adds toxicity and prompt injection. Pro adds PII and factual accuracy. Enterprise adds brand safety. Enterprise limits, retention, SSO, SLA, and any BAA are confirmed in your agreement.
One inference event submitted through an SDK counts as one logged record. One agent run also counts as one logged record; the steps stored within that run do not add logging units. If you log both an inference event and its agent run, each record counts once. Repeating an identical event or run ID does not count it twice.
Classification analyzes an inference event using the detector types available on your plan. It has its own monthly allowance, separate from logging. Once that allowance is exhausted, additional analysis waits for available capacity. Logging overages do not purchase more classifications.
Yes. Event logging, AI classification, and playground allowances reset monthly on both payment schedules. Annual subscriptions charge $996 for Growth or $4,992 for Pro upfront; $83 and $416 are the monthly equivalents. Your usage window and payment renewal date are separate.
Starter stops new logging at 10,000 records. Active Growth and Pro subscriptions support additional logging at the listed rate; overages are calculated and invoiced separately for each monthly usage window, including on annual plans. Trial logging is capped at the included allowance. Usage alerts are checked daily at 50%, 80%, and 100%.
New paid subscriptions use secure checkout. Existing subscribers can manage billing through the customer portal. A downgrade requested through the billing API is scheduled for the end of the current paid period, so your current access remains until then. Enterprise changes are arranged with sales.
First-time paid subscribers are eligible for a 14-day trial, including customers who already have a billing account but have never subscribed. Checkout shows the trial and first charge before you confirm. Starter remains free without a payment card.
The hosted playground has the message allowance shown above. Your own application and agent executions use your model-provider account; those provider charges are separate from DriftRail logging. Access to evaluations, caching, or simulations does not promise unlimited model execution.
Contact us to discuss eligibility and available promotions. Any approved discount is shown at checkout or in your agreement.
Start with 10,000 logged records, 1,000 AI classifications, and an agent workspace. No credit card required.
Get started free