Simple, transparent pricing
One flat base fee. A pooled evaluation quota for your whole team. Seats included — add more anytime.
Starter
For small teams getting started
- 5,000 evaluations/month (pooled)
- 3 seats included, +$7/seat after
- All safety gates included
- Audit trail (tamper-proof, exportable)
- Connect ChatGPT, Claude, Copilot
- Overage: $1.00/1K evals
Pro
For teams with compliance requirements
- 100,000 evaluations/month (pooled)
- 10 seats included, +$7/seat after
- Everything in Starter, plus:
- Human approval workflows (EU AI Act Art 14)
- Webhooks + real-time alerts
- 10 pipelines, workspaces, channels
- Overage: $0.25/1K evals
Enterprise
Priced per account · annual invoicing
For regulated organisations
- 1,000,000 evaluations/month (pooled)
- 25 seats included, +$5/seat after
- Everything in Pro, plus:
- Custom safety rules + device policies
- Emergency killswitch (device/site/global)
- SSO (SAML/OIDC) + 2FA enforcement
- Unlimited pipelines, workspaces, channels
- Invoice billing · Overage: $0.10/1K evals
Cost Calculator
Base fee + extra seats + overage beyond your pooled quota — across every plan, side by side
Live plan constants from /api/public/plans — what you see here is what billing uses.
What will I use?
Real benchmarks from production integrations
| Use Case | Evals/mo | Seats | Est. Cost | Plan |
|---|---|---|---|---|
| Solo dev with Claude Code + MCP proxy | ~200 | 1 | $99/mo | Starter |
| LangChain agent with shell access | ~2,000 | 2 | $99/mo | Starter |
| CrewAI multi-agent workflow | ~8,000 | 4 | ~$109/mo | Starter |
| MSP managing 50 network devices | ~15,000 | 6 | $299/mo | Pro |
| Game NPC engine (500K+ interactions) | ~80,000 | 12 | ~$313/mo | Pro |
| Casino AI bots (regulated) | ~200,000 | 15 | Custom | Enterprise |
| Enterprise AI platform (multi-tenant) | ~1,500,000 | 30 | Custom | Enterprise |
Evaluation quotas are pooled across your whole team. Overages only on excess beyond your plan quota.
Frequently asked questions
What counts as an evaluation?
Every call to POST /v1/evaluate or task submitted to a pipeline counts as one evaluation. All gates (1-4) together count as a single evaluation.
How does pooled usage work?
Your evaluation quota is shared by your whole tenant — every seat, agent, and API key draws from one pool. No per-person limits to juggle. Seats are people with dashboard access: Starter includes 3, Pro 10, Enterprise 25. Extra seats are $7/mo ($5 on Enterprise).
What are the 4 safety gates?
Gate 1: Regex denylist — 36 patterns, <1μs, all plans. Immutable base laws cannot be weakened.
Gate 2: LLM judge — stateless intent evaluation, all plans. Per-pipeline endpoint (your model, A2A safety prompt). Full prompt+response logged for audit.
Gate 3: Behavioral — 5 attack patterns, fail-closed. All plans.
Gate 4: Scope enforcement — per-agent permissions (architect/expert/observer), all plans with workspaces.
All 4 gates are included on every plan. The difference is volume (5K / 100K / 1M pooled evaluations), seats, and advanced features (custom rules, device policies, SSO).
What happens if I exceed my plan quota?
Nothing breaks — evaluations keep flowing and the excess is billed as overage on the 1st of the next month. Rates per plan: Starter $1.00/1K, Pro $0.25/1K, Enterprise & Institutional $0.10/1K. The calculator shows when upgrading beats paying overage.
Can I self-host the LLM judge?
Yes. Configure your own LLM endpoint on your pipeline for Gate 2. A2A provides the safety prompt — you provide the inference. Gate 1 regex always runs locally. Gate 3 runs entirely on your own infrastructure. Zero cloud dependency for air-gapped deployments.
Do you store my commands?
Evaluations are logged for your OCSF audit trail but never used for training or shared. You own your data. Export JSON (SIEM) or CSV (compliance) anytime.
What compliance frameworks are supported?
EU: EU AI Act (Articles 9, 12, 14, 15), GDPR. US: NIST AI RMF 1.0, Colorado AI Act, California SB 53/942, Texas TRAIGA. Industry: SOC 2, HIPAA, PCI DSS, ISO 27001. See EU mapping | US mapping.
