You've been burned before.
The last agent tool you bought charged priced after a short conversation before you ran your first call.
This one runs free until a sealed ProofPacket proves it's worth paying for.
The invoice is the proof.
"ProofPacket-sealed accounting is not a feature. It's the invoice."
Builder computes. Verifier refutes — on independent, read-only access to the same run, never trusting the Builder's own claim of success. Every billable event, a routed call, an eval, a gate-run, gets hash-chained into that same sealed ProofPacket ledger your engineering team already audits. There's no separate billing system to take on faith: the invoice resolves to the identical chain that proves the work happened.
Four tiers. Zero surprises.
Human seats are unlimited and free on every plan — you pay per agent, never per person.
Broad capability that works out of the box. Same Builder ≠ Verifier runtime, no vertical tuning required.
Tuned per vertical or role, with domain memory and priors loaded. Eligible agents carry a gold tuned badge.
Built on request for a specific team's stack. Scoped contract, dedicated verifier fleet, and onboarding.
- Full Builder + Verifier runtime — nothing crippled
- ≤5 forecasters · 100 calls · 1K gate-runs / week
- ProofPackets sealed per run, truncated (7-char hash visible, chain withheld)
- Weekly drift summary
- Support: community
- Same full runtime
- Real per-agent limits — 10× free tier
- ProofPacket JSON export
- Real-time drift alerts
- Email + Slack support · 48h
- Dedicated verifier compute
- Full hash chain — public verify endpoint
- SOC 2 audit-ready docs
- Multi-team workspaces · SSO / SAML
- 8h support · 99.5% SLA
- Same runtime, on your hardware (VPC / bare metal / air-gapped)
- Signed packets chained to your ledger
- HIPAA / FedRAMP-ready
- 24/7 support · 99.95% SLA
Domain Expert pricing is per-agent: priced after a short conversation (eligible agents priced after a short conversation) — see per-agent pages.
The full comparison.
If a capability isn't listed here, it isn't tier-gated.
| Feature | Free | Team | Pro | Governance |
|---|---|---|---|---|
| Builder + Verifier runtime | ✓ | ✓ | ✓ | ✓ |
| ProofPackets sealed per run | ✓ truncated | ✓ | ✓ | ✓ |
| JSON export | — | ✓ | ✓ | ✓ |
| Full hash chain · public verify endpoint | — | — | ✓ | ✓ |
| Weekly drift summary | ✓ | ✓ | ✓ | ✓ |
| Real-time drift alerts | — | ✓ | ✓ | ✓ |
| Multi-team workspaces | — | — | ✓ | ✓ |
| SSO / SAML | — | — | ✓ | ✓ |
| Dedicated verifier compute | — | — | ✓ | ✓ |
| SOC 2 audit-ready docs | — | — | ✓ | ✓ |
| SLA | — | — | 99.5% | 99.95% |
| On-prem / VPC deployment | — | — | — | ✓ |
| Packets chained to your ledger | — | — | — | ✓ |
| HIPAA / FedRAMP-ready | — | — | — | ✓ |
| Support response | community | 48h | 8h | 24/7 |
Thirty-one agents. Thirty-one line items.
Team-tier price per agent, free-tier weekly limit, and the metered unit where one applies. Metered agents carry no flat fee — you pay only for what clears verification.
| Agent | Installs | Free limit / week | Team | Metered unit |
|---|---|---|---|---|
| priorv2.4.1forecasting calibration · flagship | 12.4k | 5 forecasters | Priced after a short conversation | — |
| judgev1.9.0agent evaluation | 8.1k | 100 evals | Priced after a short conversation | — |
| gatev1.4.7proof-of-done | 6.7k | 1K gate-runs | Priced after a short conversation | — |
| routev3.1.2LLM routing | 9.9k | 1K routed calls | usage-based | Priced after a short conversation/ 1K tokens |
| harvestv0.8.3competitive intel | 5.3k | 50 signals | Priced after a short conversation | — |
| guardv2.0.0determinism gates | 7.8k | 100 replays | Priced after a short conversation | — |
| draftv1.2.5GTM personalization | 4.6k | 25 emails | usage-based | Priced after a short conversation/ email |
| meshv0.6.1multi-agent orchestration | 11.2k | 500 envelopes | Priced after a short conversation | — |
| benchv0.3.0edge-inference throughput bench | 1.4k | 50 bench-runs | Priced after a short conversation | — |
| routerbenchv0.4.1multi-LLM routing benchmark | 2.1k | 50 tasks | Priced after a short conversation | — |
| harnessv0.5.0coding-agent harness evolution | 1.9k | 20 evolution-runs | Priced after a short conversation | — |
| mregtestv0.2.4replay-based regression testing | 2.6k | 50 replays | Priced after a short conversation | — |
| evalobsv0.3.2fleet SLO + anomaly observability | 1.7k | 500 telemetry windows | Priced after a short conversation | — |
| operatorv0.4.0plan-execute-verify ops loop | 2.3k | 100 objectives | Priced after a short conversation | — |
| memoryadmitv0.2.1adaptive memory admission control | 1.2k | 1K admissions | Priced after a short conversation | — |
| mcruntimev0.3.5model-check + runtime verification | 0.9k | 200 model-checks | Priced after a short conversation | — |
| deviatev1.0.0deviation-engine artifact scoring | 3.1k | 500 scores | Priced after a short conversation | — |
| monotonicv0.2.0monotonic risk correlation | 0.8k | 1 monitoring cycle | Priced after a short conversation | — |
| edgeshardv0.5.3collaborative edge/cloud sharding | 2.8k | 1K shard-plans | usage-based | Priced after a short conversation/ 1K tokens |
| ltvv0.4.4acquisition/retention value split | 1.5k | 10 cohort runs | Priced after a short conversation | — |
| synthv0.1.8synthetic ground-truth generation | 1.1k | 1K synthetic records | Priced after a short conversation | — |
| rig-routev0.6.0LLM gateway | 1.6k | 1K routed calls | Priced after a short conversation | — |
| rig-benchv0.2.7fleet benchmark orchestration | 0.5k | 50 bench-runs | Priced after a short conversation | — |
| rig-graderv0.3.1rubric grading | 0.9k | 100 grades | Priced after a short conversation | — |
| rig-evolverv0.4.2harness evolution | 0.7k | 20 evolution-runs | Priced after a short conversation | — |
| rig-replayv0.3.0replay determinism | 1.1k | 50 replays | Priced after a short conversation | — |
| rig-metricsv0.4.0eval observability | 1.3k | 500 telemetry windows | Priced after a short conversation | — |
| rig-captainv0.5.1strategy-switch operator | 0.6k | 50 objectives | Priced after a short conversation | — |
| rig-boundedv0.3.3memory admission | 0.8k | 1K admissions | Priced after a short conversation | — |
| rig-verifier-runtimev0.6.2model-check + runtime verification | 0.4k | 200 model-checks | Priced after a short conversation | — |
| rig-deviantv0.4.5deviation scoring | 1.0k | 500 scores | Priced after a short conversation | — |
Asked by every finance team.
Six questions, short answers, no asterisks.
Paste it into the request.
Written for the person who approves the spend, not the person who runs the eval.
RIG replaces manual agent QA with a sealed verification trail: a 4-agent Team deployment costs priced after a short conversation at priced after a short conversation — priced after a short conversation at the first-100-slot rate of priced after a short conversation — against ~6 eng-hours/week of manual review, roughly $15,600/month at a $100 loaded hourly rate. Every gate run, eval, and routing decision is metered inside its own ProofPacket, so finance gets a hash-chained invoice trail, not a usage estimate. The pilot runs on the free tier, so the evaluation costs priced after a short conversation and produces the audit evidence either way.
The first 100 slots are already moving.
First 100 install slots lock Team tier at priced after a short conversation for the life of the subscription — after that it's the standard priced after a short conversation. 62 remaining, read directly from the sealed install ledger: the same hash-chained runtime that seals your ProofPackets, not a marketing counter. No countdown clock — the number only moves when a slot installs and seals.
The pilot is free.
Free tier, full runtime, sealed receipts from the first run.