Pay for the verdict, not the telemetry.
Capture is free forever — the OSS local tool and a hosted free tier never get gated. You pay only when a shared trust record, alerting, or enforcement starts to matter. No checkout on this page: every plan starts from signup.
Plans
OSS Local
The independent referee, on your machine. Code never leaves the box — only structural metadata is diffed.
- Deterministic claim-vs-reality diff
- All hooks: Claude Code, Cursor, Codex, CI
- JSONL stored on your machine
- Self-host, no account required
Free Cloud
Hosted single-user dashboard so you can see your first catch land — free-forever, not a trial.
- Everything in OSS Local
- Hosted single-user dashboard
- Deterministic verdicts (no judge, no gate)
- Hard cap at 500 runs — no overage
Team
The moment a second developer's agent activity needs to live in one shared trust record. Flat, no per-seat math.
- Everything in Free Cloud
- Cross-developer team dashboard
- Slack / email alerting
- Deterministic verdicts across all vendors
- LLM judge available as opt-in add-on
Business
Stop observing, start enforcing. The policy gate turns agentwatch into an active gatekeeper on your merges.
- Everything in Team
- Policy gate — CI merge-block on false claims
- Cross-vendor trust leaderboard
- Exec / leadership trend view
- Judge add-on included at cost+
Enterprise
The neutral cross-vendor record as governance infrastructure — for security, compliance, and self-host.
- Everything in Business
- SSO / SAML & SCIM
- Self-host / on-prem Helm chart
- Audit export + EU AI Act Art. 19 packaging
- SLA, dedicated compute, custom contracting
All prices in USD. Team and Business bill flat per org, per month — no per-seat math. Overage is silent and proportional ($5 / 1K runs on Team, $4 / 1K on Business); you are never charged for usage you didn’t have. The LLM judge is a separately metered pass-through add-on so it never erodes your flat tier. Enterprise pricing (~$30K–60K/yr) is directional and negotiated per deal.
What unlocks where.
The same usage axis runs across all five tiers. Retention is the cleanest paid lever (30 days → 90 days → 1 year → custom); the policy gate is the Business step; SSO and self-host land at Enterprise.
| Feature | OSS Local | Free Cloud | Team | Business | Enterprise |
|---|---|---|---|---|---|
| Plan | |||||
| Price | $0 forever | $0 forever | $49/org/mo | $299/org/mo | Custom |
| Included runs / mo | Unlimited (local) | 500 | 5,000 | 25,000 | Committed / unlimited |
| Overage | — | none (hard cap) | $5 / 1K runs | $4 / 1K runs | Negotiated |
| Users | 1 (local) | 1 | Unlimited | Unlimited | Unlimited + SCIM |
| Retention | Local only | 30 days | 90 days | 1 year | Custom / unlimited |
| Capture & verdicts | |||||
| Deterministic diff | ✓ | ✓ | ✓ | ✓ | ✓ |
| Cross-vendor capture (Claude Code / Cursor / Codex / CI) | ✓ | ✓ | ✓ | ✓ | ✓ |
| Hosted dashboard | — | Single-user | Team | ✓ | ✓ |
| Alerting (Slack / email) | — | — | ✓ | ✓ | ✓ |
| LLM judge | — | — | Add-on | Incl. at cost+ | ✓ |
| Enforcement & governance | |||||
| Policy gate (CI merge-block) | — | — | — | ✓ | ✓ |
| Cross-vendor trust leaderboard | — | — | — | ✓ | ✓ |
| Exec / leadership trend view | — | — | — | ✓ | ✓ |
| SSO / SAML / SCIM | — | — | — | — | ✓ |
| Self-host / on-prem (Helm) | self-host OSS | — | — | — | ✓ |
| Audit export / EU AI Act packaging | — | — | — | — | ✓ |
| SLA | — | — | — | — | ✓ |
The questions teams ask first.
When do I actually start paying?
What counts as a run?
Is the local tool really free?
Claude Code, Cursor, Codex, CI), JSONL stored on your own machine. Your code never leaves the box; only structural metadata is diffed. The local tool is the independent referee, not a loss-leader we plan to gate later.