The execution layer
for production
AI agents.
Policy before every action. Exactly-once execution guarantee. SHA-256 receipt on every outcome.
Free to start · No credit card required · Self-hostable
Press ⌘K to search docs
Works with your entire stack
Agents in production
need a guardrail layer.
Your agent just charged a customer twice. There’s no receipt, no audit trail, no way to know which policies ran or why. You find out when support calls.
Every AI agent that touches production systems — billing, CRM, communication, data — needs policy evaluation before it acts and a tamper-evident record of what happened after.
See everything. Control everything.
Every agent action observable, every policy decision auditable, every connector under one roof.
OBSERVE
GOVERN
CONNECT
Home
Last 24hActions Today
12,847
+1,203 vs yesterday
Execution Rate
99.97%
of committed
Escalations
3
awaiting review
Receipts Minted
1.28M
all time
Recent Actions
↻ auto-refresh
| ID | Service | Action | Status | Time |
|---|---|---|---|---|
| act-f93a | svc-landing | deploy_landing_page | COMPLETED | 4s ago |
| act-e41b | svc-api | run_test_suite | COMPLETED | 12s ago |
| act-d28c | svc-infra | update_dns_record | ESCALATED | 1m ago |
| act-c77d | svc-comms | send_slack_alert | COMPLETED | 3m ago |
| act-b19e | svc-api | scale_api_workers | COMPLETED | 7m ago |
Three lines to
governed execution.
Propose an action. The policy engine evaluates it against your rules. Execute with an exactly-once guarantee. Every outcome gets a tamper-evident receipt.
123456789101112131415from statis_ai import StatisClientclient = StatisClient(api_key="sk-statis-...")# Propose an action — nothing executes yetaction = await client.propose(entity_id="acct-8821",action_type="apply_discount",payload={"percent": 15, "reason": "churn_risk"},)# Policy engine evaluates. Execute if approved.if action.status == "APPROVED":receipt = await action.execute()print(receipt.id) # sha256:a3f29c...❯
How we think about governed execution.
Five principles that shape every decision in the Statis codebase. These aren't marketing — they're the trade-offs we refuse to make.
Determinism over ML.
Your governance layer shouldn't hallucinate. Rules are versioned, testable, reversible — no magic, no prompts in the critical path.
Audit is the product.
The ledger isn't a feature, it's the thing you're paying for. Every receipt tamper-evident, queryable, and exportable.
Operator-first tooling.
SDKs, CLIs, and infrastructure-as-code. No required dashboard. Built for the people who actually own production.
Self-hostable by default.
Docker Compose, bring your own database, run on your own metal. No vendor lock-in on the trust layer.
Reversible by design.
Every policy versioned, every decision explainable, every action undoable. Mistakes should be recoverable, not catastrophic.
Agents need
infrastructure,
not guardrails.
Every action flows through a deterministic trust layer. Every outcome is receipted. Every decision is reversible, auditable, and reproducible. That's the product.
Frequently asked.
We used Statis to build Statis.
Every action proposal, policy check, and execution receipt on this site ran through our own infrastructure.
Enterprise-ready from day one.
The controls your security team will ask for before procurement — already built in.
SOC 2 Type II
Audit in progress. Security controls across access, availability, and confidentiality.
OIDC SSO
Native Okta, Entra ID, and any OIDC-compliant provider. SCIM provisioning included.
Self-hostable
Deploy on your own infrastructure with Docker Compose. No data leaves your VPC.
Tamper-evident Receipts
Every action produces a SHA-256 signed receipt. Immutable. Auditable. Cryptographically verifiable.
Data Residency
Choose your region. On-prem deployment available for regulated industries.
Full Audit Logs
Complete decision history. Every policy evaluation, every escalation, every outcome — forever.