EVE — Positioning & Category (16D)
Central positioning statement (approved)
EVE governs AI execution before side effects occur and produces independently verifiable evidence of what was approved, what executed, and why.
What EVE is
EVE is an independent governance runtime that makes deterministic decisions before governed execution, cryptographically binds authorization to the resulting action, and produces evidence that can be verified separately from the system that generated it.
Supporting distinctions (approved, factual)
| Others | EVE |
|---|---|
| Prompt filters inspect content. | EVE makes deterministic execution decisions (ALLOW / BLOCK / MODIFY) before a tool runs. |
| Agent-security tools detect suspicious behavior. | EVE enforces a decision at the execution boundary and emits signed governance evidence. |
| Logging records what a system says happened. | EVE Proof lets a third party independently verify governance artifacts without trusting the EVE server. |
| Model-provider guardrails operate inside one provider. | EVE is designed as an independent governance layer across models, agents, tools, and MCP systems. |
Not
EVE is not merely a scanner, a dashboard, or a logging layer. It is designed to make deterministic decisions at the execution boundary, bind the approved request to the executed one, and produce verifiable evidence — but it is not a compliance certification, does not guarantee regulatory outcomes, and (per the claims registry) does not provide remote exactly-once execution or full RFC 8785 float compliance.
Readiness (see readiness.md)
EVE SDK core is a RELEASE CANDIDATE WITH EXCLUSIONS. The governance architecture is pilot-validated. This is not a production-readiness claim.
Limitations
Framework adapters other than the generic one are experimental; the pip client governs via
hosted mode; jcs-1 is a constrained canonicalization profile; cooperative hooks are not an
unbypassable boundary. Full limitations are in each product page and the claims registry.