Skip to content
bucker

Pricing

The price is on the page

Flat monthly platform plans with unlimited human seats, plus one metered lane that bills only for fixes a sandbox verified and a human approved. Prices are read from GET /billing/plans — the same endpoint the product itself reads when it enforces your quota — so what is published here and what you are charged cannot drift apart.

Plans

Every plan publishes a monthly fair-use ceiling on the model tokens and sandbox minutes used for AI work that is not a verified fix (summaries, explanations, investigations, cluster labels, forecasts). Past it, you get the deterministic result at once and the model run is queued for the next UTC day. Nothing is billed for it and nothing is dropped silently. A verified fix you have approved is never queued by the ceiling.

All prices in US dollars, billed per calendar month. Enterprise is quoted rather than self-serve. The 14-day Pro trial needs no card and is available once per organization; if you do not convert it lapses to Free by itself and nothing is charged.

What we promise about your bill

Bill shock is this category’s defining resentment: a bad deploy that triples the invoice, credits that expire at the end of the month, a per-seat charge that grows with the team. Each of the promises below is a mechanism in the code rather than a sentiment, and each one names the file that implements it.

Your worst day is on us
A storm from one bad deploy is one issue. Over-quota events cost nothing — they degrade rather than bill — and because the AI lane meters distinct issues, a million-event storm can produce at most one billable fix. This is arithmetic in billing/quota.ts, not a goodwill policy: the storm estimator returns a literal zero for event overage, and a test asserts it.
AI billed only on verified outcomes
A fix is billable only if it reached verification tier A or B and a named human approved it. A rejected fix, a failed sandbox run and an unreviewed proposal are all free. Fuzzy outcome definitions turn into invoice disputes, so the rule is a pure function — billing/billability-assessment.ts — that you can read.
No per-seat AI tax
Human seats are unlimited on every tier, including Free. There is no seat count field in the plan schema at all — the type is the literal string "unlimited", so a per-seat limit could not be added without changing the type, and the test suite asserts that no perSeatCents or maxUsers field has appeared.
No SSO tax, no per-unit upsells
Single sign-on and SCIM are on every paid tier, not held back for Enterprise. Cron monitors, uptime monitors, user feedback and session replay are included on every tier including Free. Both are enforced by tests that fail the build if a flag is ever turned off.
Credits roll over
Unused verified-fix credits carry into the next month instead of expiring. The honest caveat: the balance is capped at three months of your allotment, so credits do not expire but they also do not accumulate forever. That cap is ROLLOVER_CAP_MONTHS = 3 in the billing module, and rollover credits are always spent before you are metered.
The allowance is in units, and an error is one
Every plan’s volume is counted in weighted units, the one number the quota, your usage page and the 80, 90 and 100 percent notices all read. An error is 1 unit; a transaction or a profile is 0.2; a log record, a product event or a metric data point is 0.05; a synced source row is 0.01; sessions and feedback are free. So 100,000 log lines use 5,000 of your units, not 100,000 “events”. The weights are a published register, not a meter: there is no per-category price and no separate invoice line for any of them. The usage page shows both the units and the raw item count, and says which one the allowance is measured in.
Caps are on by default, and drops are receipted
Every tier ships with spend caps enabled. Past your unit allowance, ingestion degrades to stratified sampling rather than dropping traffic silently, and every dropped event is accounted for in a “what was dropped and why” panel in your dashboard. The AI budget gate refuses to start a run that would exceed a hard cap, rather than refusing to bill for work already delivered.

Two things to know before you compare

Talk to us

Enterprise is quoted. Tell us what you need — a signed DPA, an annual term, a security review — and a person reads it. The message goes to the operator console of this deployment, not to a marketing tool.

What is on every plan

Features included on each plan
FeatureFreeProTeamEnterprise
Single sign-on (SAML / OIDC)Not includedIncludedIncludedIncluded
SCIM directory sync, including agentsNot includedIncludedIncludedIncluded
Audit APINot includedIncludedIncludedIncluded
Custom rolesNot includedNot includedIncludedIncluded
SIEM streamingNot includedIncludedIncludedIncluded
Break-glass accessNot includedNot includedNot includedIncluded
Cron monitorsIncludedIncludedIncludedIncluded
Uptime monitorsIncludedIncludedIncludedIncluded
User feedbackIncludedIncludedIncludedIncluded
Session replayIncludedIncludedIncludedIncluded

Asked for, and not sold

These appear on most price lists in this category, or would read as one from the flag alone. They are not on ours, on any tier, because the deployment cannot honour them today — not as a packaging decision.

EU region
We operate one region. The EU control plane is marked inactive in the residency registry and is not deployed, so nothing runs in it.
Data processing agreement
The DPA is published as an outline, not a signable contract: no sub-processor is contracted and no breach-notification window is committed.
All integrations
Every integration already connects on every tier, Free included — there is no plan check in the product, so this cannot honestly gate on Team and up.