Pricing
Deterministic enforcement costs for production automations. Measured in protected actions.
All plans enforce the same core safety rules. Pricing affects request volume, retention duration, analytics depth, and support level — not enforcement strictness.
WARN approvals are available only on plans that include human approval capabilities. Plans without approvals fail-closed.
AI Safety Gate pricing
How billing works
One validation call before an action counts as one protected action.
Core enforcement behavior is consistent across tiers: PASS allows execution, WARN pauses for review (if available on your plan), and BLOCK stops execution. Plans change quotas, retention, analytics, and support — not the core safety outcome semantics.
COUNTS AS 1
validate()
A single request producing PASS/WARN/BLOCK.ⓘ
BEST PRACTICE
Gate every irreversible step
Money, messaging, CRM writes, production updates.
Enterprise consideration
Audit retention scales
Higher tiers keep longer logs for investigations.
What is a protected action?
Natural-language definition for production usage.
A protected action is a workflow step guarded by AI Safety Gate. In practice, it is one call to validate() made immediately before an irreversible step (money, messaging, CRM write, production update). The returned PASS/WARN/BLOCK decision determines whether your automation continues, routes to review, or stops.
Audit retention scales with higher plans so teams can investigate incidents, demonstrate controls, and support operational requirements without changing enforcement behavior.
FREE
50 protected actions / month
$0
/ mo
Rule complexityBasic rules
TeamsLocked
Managed Safety ModeAlways enforced for safety
Logs & auditing3 days
WARN handlingWARN handling: Auto-block
Charts + analyticsLocked
SupportDocumentation only
Starter
2,000 protected actions / month
$29
/mo
Rule complexityBasic rules
TeamsLocked
Logs & auditing14 days
WARN handlingWARN handling: Auto-block
Managed Safety ModeAlways enforced for safety
SupportDocumentation support + support chat
Charts + analyticsLocked
For evaluation and early testing.
Get startedPro
Recommended
10,000 protected actions / month
$99
/mo
Rule complexityAdvanced rules + approvals
TeamsLocked
Logs & auditing90 days
WARN handlingWARN approval (approve / reject)
Managed Safety ModeConfigurable (can be disabled)
SupportPriority support
Charts + analyticsIncluded
EnterpriseComing Soon
50,000 protected actions / month
Contact
/mo
Rule complexityAdvanced rules + approvals
TeamsIncluded
Logs & auditing365 days
WARN handlingWARN approval (approve / reject)
Managed Safety ModeConfigurable (can be disabled)
SupportPriority support
Charts + analyticsIncluded
Fail-closed enforcement
No prompt-only safety
Auditable decisions
Human approval on WARN
Compare plans
Feature access is enforced server-side. If in doubt, the system defaults to the safest posture.
| Feature | Free | Starter | Pro | EnterpriseComing Soon |
|---|---|---|---|---|
| Managed Safety Mode | Always enforced for safety | Always enforced for safety | Configurable (can be disabled) | Configurable (can be disabled) |
| Teams (members, roles, invites) | Locked | Locked | Locked | Included |
| Rule editing (Policy Control Mode) | Locked | Locked | Included | Included |
| Rule match details | Locked | Locked | Locked | Locked |
| WARN approval (approve / reject) | Auto-block | Auto-block | Included | Included |
| Fail-closed on timeout | Included | Included | Included | Included |
| Charts + analytics | Locked | Locked | Included | Included |
| Audit retention duration | 3 days | 14 days | 90 days | 365 days |
Pricing logic
Infrastructure pricing should be deterministic and auditable.
Protect what matters
You pay for enforcement events—every time AI output is evaluated before an action.
Audit depth by tier
Retention increases as stakes rise so teams can investigate and demonstrate controls.
In v1, retention is enforced at the query/visibility level; data is not physically deleted. See Retention.
Operational support
Higher tiers add tighter response times and onboarding support for critical workflows.