Tracing, cost, and evaluation come with every plan. You pay for how much you store, how long you keep it, and how many policies you run. No per-seat charges.
Try it on a real agent
Runtime policies and approvals
More agents, more controls
Governance and compliance at scale
| Compare plans | Free$0/month | Team$299/month | Business$799/month | EnterpriseCustom |
|---|---|---|---|---|
| Usage | ||||
| Events per month | 50K | 2M | 10M | Custom |
| Data retention | 7 days | 90 days | 1 year | Custom, with legal hold |
| Users | 3 | Unlimited | Unlimited | Unlimited |
| Workspaces | 1 | Unlimited | Unlimited | Unlimited |
| Observe | ||||
| Traces and sessionsSDK, OpenTelemetry, Claude Code, Cursor | ||||
| Cost attributionBy agent, model, trace, and workspace | ||||
| Agent dashboards and metrics | ||||
| Guardrail posture | ||||
| Evaluate | ||||
| Prompt registry and deploy labels | ||||
| Prompt playground | ||||
| Datasets | ||||
| ScorersBuilt-in, LLM-as-judge, code, Jev Decision | ||||
| Experiments and comparison | ||||
| Govern | ||||
| Active policiesSpend Cap, Refund Guard, Loop Cap, and the rest | 1 | 3 | 15 | Unlimited |
| Decision Log | ||||
| Simulate on past traffic | ||||
| Human approval for refunds and purchases | — | |||
| Policy email alert recipients | 1 | 5 | 20 | Unlimited |
| GatewayEnforce without the SDK | — | — | ||
| Prove | ||||
| Compliance HubAI system registry, reviews, incidents | — | — | — | |
| Audit log | — | — | — | |
| Evidence packs | — | — | — | |
| EU AI Act and HIPAA modules | — | — | — | |
| Support and billing | ||||
| Support | Community | Priority email | Dedicated | |
| Billing | No card needed | Card, monthly | Card, monthly | Invoice or AWS Marketplace |
Unlimited users on every paid plan. Invite engineering, product, and risk without counting heads.
Going over your events never switches a policy off. Spend Cap and Refund Guard keep working while we sort out your plan.
Prompts, datasets, scorers, and experiments are included from Free up. Bring your own model keys.
One span that Traccia stores: an LLM call, a tool call, or an agent step. A run with 12 spans is 12 events. Traces are not counted separately, and policy decisions, approvals, and alert emails never count.
Nothing stops. Traccia keeps accepting your traces and your policies keep enforcing. We will reach out to move you to the right plan. There are no automatic overage charges.
Any policy you switch on, whether it stops a call before it runs or alerts after a run. Drafts and Simulate runs do not count, so you can test a policy against last week's traffic before you spend a slot on it.
No. An active policy blocks, reshapes, or alerts the same way on every plan. Plans change how many policies you can run, not how well they work.
Yes. Prompts, datasets, scorers, and experiments are on every plan. The playground and LLM judges use your own provider keys, so Traccia never marks up model tokens.
Yes. Start a 14-day trial with full Enterprise access and no card. When it ends you return to Free unless an admin subscribes. Have a coupon? Redeem it from Billing & Plans. How to redeem a coupon
No. The Python and TypeScript SDKs are open source and can export to Jaeger, Tempo, or any OpenTelemetry backend without a Traccia account.
Questions about pricing? We're here to help.