NULLRUNCompare
What competitors do.
What NullRun does.
Every vendor is scored on the same four axes:
gate, approval, spend, and audit.
A cell reflects capability that is publicly documented today. Roadmap promises, slide-deck screenshots and "we have a beta" do not count.
Capability comparison
why this is public
Honest comparison.
We publish this because the AI-governance category is genuinely confusing right now. Every vendor here is scored on the same four axes, so you can decide whether the gap between us is real, or marketing.
YesPartial / complementsNot supported
| Vendor | Gate | Approval | Spend | Audit |
|---|---|---|---|---|
| Direct decision-layer competitors | ||||
NullRun Runtime authorization for AI agents. | Yes | Yes | Yes | Yes |
Microsoft AGT | Yes | Partial | Not supported | Yes |
APort | Yes | Partial | Partial | Yes |
Credo AI | Partial | Yes | Not supported | Yes |
Straiker | Partial | Not supported | Not supported | Yes |
Isonapse | Yes | Partial | Partial | Yes |
| Adjacent — gateways, security, advisory | ||||
Portkey | Yes | Partial | Yes | Yes |
Noma Security | Yes | Partial | Not supported | Yes |
Prisma / Palo Alto | Partial | Not supported | Partial | Partial |
Advisory shops | Not supported | Not supported | Not supported | Not supported |
As of 26 October 2026. See our security page for our controls, and our trust page for what we publish about NullRun.
The runtime decision point is the product.
The rest is plumbing.
Every other surface in this category — prompt recorders, model guardrails, observability dashboards, advisory decks — answers a different question.
An autonomous agent forces this question: is this action authorized, right now, by the policy that exists today, before it runs?