A Policy Decision Point You Can Wire In
You do not want another dashboard - you want an endpoint. SentriAI gives you a stateless, deterministic Policy Decision Point you call per action from an API gateway, MCP middleware, a LangChain hook, a sidecar, or an in-process SDK.
What you get
As a security engineer you build the controls the org relies on, and you want them deterministic, testable, and low-latency. SentriAI’s decision layer is a pure function: the same action envelope always yields the same verdict, risk, trust score, and ledger hash. You wire it in wherever it fits, dry-run policy changes safely, and get a fail-soft contract that never returns a 500.
- A stateless fast-path PDP you call per action from any enforcement point
- Deterministic verdicts - the same envelope always yields the same result
- A dry-run path to test policy changes without writing events
- Fail-soft by contract: malformed input returns a safe verdict, never a 500
Mapped to what matters to you
| Your priority | How SentriAI helps |
|---|---|
| Integrate without lock-in | One call shape works as a gateway, MCP middleware, hook, sidecar, or SDK |
| Test policy safely | Dry-run returns the verdict without persisting an event |
| Keep it fast | The PDP is stateless and skips the heavy investigation on the fast path |
| Trust the output | Deterministic, reproducible verdicts and ledger hashes you can assert on |
| Never break the caller | Fail-soft: an ambiguous action resolves to human review, not an error |
Where it fits your work
How a security engineer uses SentriAI
- Put the MCP guard or gateway in front of the PDP as a policy enforcement point
- Assert on deterministic verdicts and ledger hashes in tests as golden vectors
- Use dry-run in CI to confirm a policy change produces the expected verdicts
- Read active org-scoped policies and the engine summary from the API
Why SentriAI
Frequently asked questions
How do I integrate SentriAI as a security engineer?
Pick the enforcement point closest to your agents - an API gateway, MCP middleware, a LangChain hook, a sidecar, or an in-process SDK - and have it POST an action envelope to the Policy Decision Point and apply the verdict. The same call shape works everywhere.
Is the decision deterministic?
Yes. The PDP is a pure function: the same action envelope always yields the same verdict, risk, trust score, band, enforcement posture, and ledger hash, with the hash computed over the decision rather than the timestamp, so it reproduces.
Can I test policy changes without side effects?
Yes. The dry-run path returns the verdict without persisting an event, so you can replay production-shaped traffic against a candidate policy in CI or staging and confirm the verdicts before promoting it.
What happens on malformed input?
The PDP is fail-soft. A malformed or unrecognized envelope returns a conservative human_review_required verdict rather than an error, so the calling enforcement point never receives a 500.
Stay in the loop
One practical finance briefing a week - new guides, checklists, and benchmarks.
Wire governance into the action path
Call one deterministic decision endpoint per action. Start free, no card required.
Talk to us