Questions skeptical evaluators actually ask

Two separate problems. Conduct is what your agent may say and do — the lines it won't cross, set to your use case. Persona is how it sounds once it's allowed to speak — softer or firmer, formal or familiar. Conduct decides what may be said; persona decides how it sounds. Decorum solves both, separately, on your own agent loop.

Prompts ask; nothing makes the model comply. Most guardrail libraries either call another model — cost, latency, another thing to jailbreak — or only check output. Decorum turns your conduct and persona into rules enforced across input, tool calls, and output, it fails closed, and it adds the axis a guardrails library doesn't — persona, on your own loop.

Not in the usual sense. Governance platforms give a risk office dashboards, model registries, and human-in-the-loop escalation. Decorum gives the builder a conduct and persona layer that enforces the lines pre-emptively, inside the agent loop you already run — the agent never escalates a decision, it simply can't cross the line. Our content-free audit trail supports a compliance review, but Decorum is a layer you install, not a platform you watch.

No, and we won't claim it. The deterministic input gate has published catch-rate bounds, and a reasoning-based screening tier is on the roadmap. What is absolute: enforced floors on output are deterministic — a rule you define as a floor is checked every turn, regardless of what the model was talked into.

The guard is text-scope: it screens what your user would see. Reasoning traces pass through unscreened, and the integration docs say so explicitly — hosts that render reasoning to users should treat that surface accordingly.

Any TypeScript or JavaScript agent loop. Five hooks wire the whole contract; reference adapters cover a zero-dependency loop and the Vercel AI SDK. The packed artifact runs on plain Node — no framework, no build step required.

Packs are testable artifacts. Each ships an executable conformance suite and a red-team corpus; a strict conformance pass means the guarantees were exercised, not assumed.

Yes. As the agent runs, the engine emits content-free attestation: what should have constrained the agent, what the runtime did at each stage, and why. A separate, independent verifier then cross-examines those claims against the observed inputs and outputs and renders ALIGNED, MISALIGNED, or INCOMPLETE — offline, deterministic, no model involved. It catches a runtime that claims an enforcement it didn't perform.

It proves conduct rules were enforced exactly as your packs specify, and that your persona was delivered to the model unmodified. It does not prove the agent behaved in character — persona adherence is a behavioral claim a deterministic check can't honestly make. We'd rather hand you a narrow verdict that stands up than a broad one that doesn't.

No. Decorum is proprietary commercial software, licensed per deployment. You receive a versioned, installable artifact and documentation.

None. There is no Decorum server — the engine runs inside your infrastructure and makes no network calls. The audit trail it writes is content-free by construction and stays with you.

Three small libraries, no network, deterministic checks. Buffered output delivery is the one latency trade, and it's your call per deployment binding.

Decorum is built by Getman AI. It governs production agents today on two live products, including a real-estate transaction copilot where information-not-advice is a liability line.