Rigorix
Rigorix gates consequential actions proposed by AI agents, rechecks operator-defined authority and conditions, and records signed decisions. Platform and risk teams use it to control execution and verify evidence for actions such as payments, migrations, and production changes.
Who it's for
- Platform engineering teams routing AI-agent actions through a controlled execution path for production changes or database migrations.
- Risk and operations teams that need authority and eligibility rechecked before financial transactions are dispatched.
- Teams that need independently verifiable evidence of agent-action decisions, including offline verification.
Not the right fit if…
- Teams needing a sandbox for agent activity: Rigorix explicitly says it is not a sandbox and only governs actions routed through its execution path.
- Teams needing cross-session capacity without a shared contested store: the page says cross-session capacity depends on sessions contending over one shared store.
How it fits your workflow
An agent proposes an operation, which is routed through a locked plan and an approval bound to the intent. The engine checks sequence, operator-defined preconditions, effect identity, and capacity before deciding whether to dispatch.
After dispatch, Rigorix records the intent and handles an unconfirmed result as ambiguous. It fences dependent actions and refuses retries until a resolver presents provider evidence.
Rigorix records signed evidence for decisions, including refusals. Teams can use a published verifier to check the chain, signature, and anchor offline.
Pricing
The vendor doesn't publish prices. Check with Rigorix directly.
Prices checked on Nov 4, 2026 from the vendor's site. They can change; confirm before you buy.
Key features
- Execution gate
- Checks consequential operations before dispatch and refuses an operation when a check is unclear or fails.
- Sequence policy
- Evaluates whether a step is allowed given the run prefix and signed history, including across runs.
- Dispatch-time preconditions
- Runs operator-authored checks immediately before dispatch. The page gives authority, eligibility, limits, and freezes as examples of conditions that can be checked.
- Canonical effect keys
- Matches actions by an operator-defined canonical effect key so that the same consequence under a different name can be identified as a duplicate.
- Aggregate capacity ledger
- Reserves capacity before authorizing an action. The in-process ledger is atomic for a single execution; cross-session capacity requires sessions to contend over a shared store.
- Commitment and settlement handling
- Records intent before sending an action and keeps an unconfirmed outcome ambiguous. Dependents are fenced and retries are refused until a resolver presents provider evidence.
Limitations to know
- A dispatch-time precondition narrows the window between checking and consequence but does not eliminate it; the downstream system still enforces eligibility at its own commit point.
- Rigorix does not know domain truth automatically. Operators provide authority checks and effect keys; the page says no model guesses what a condition such as “frozen” means in a system.
- Rigorix is not a sandbox; it governs actions routed through its execution path.
- Cross-session capacity requires every session to use and contend over one shared store. The in-process ledger provides atomicity for a single execution.
- Rigorix records the gate decision but does not vouch for the downstream system.
Alternatives to consider
-
Keydris
Choose it if you need per-action authority evaluated against versioned policies with ALLOW, APPROVAL REQUIRED, or REJECT decisions, rather than Rigorix's execution gate and multiple operation checks.
-
Execlave
Choose it if you want a managed-cloud or self-hosted AI-agent governance platform with real-time policy enforcement and immutable audit trails.
Getting started
pip install rigorix-verifier- Run the open-source authority demo.
- Inspect the signed refusal produced by the demo.
- Install a published verifier package.
- Verify the evidence bundle offline.