Cairn
Cairn
Three tiers, billed by what we actually capture for you. You can read the whole page in under a minute. We checked.
For the first prompt and the first quiet weekend of prototyping.
For teams whose agents are already in front of paying customers.
For regulated industries, self-hosted deployments, and large fleets.
One captured unit of work — a model call, a tool invocation, a planning step, an eval check. A typical agent run produces five to fifteen spans depending on tool use.
Tokens describe what the model did. Spans describe what your agent did. We bill by the second one because it's the one you can actually budget for.
Fourteen days, no card. Span limits are lifted during the trial so you can run against real traffic without rationing.
Yes, on Range. Same trace and eval surfaces as the cloud product, with your data never leaving your boundary.
No. We commit to 90 days' notice for any pricing change, applied to your next annual term — never retroactively.
Watch how a single run flows from SDK call to trace, eval, and dashboard.