REA Benchmark integrity
Claims map to behavior you can hit: INTEGRATED, cache_hit, BLOCKED_TIMEOUT, 402 at zero loops. We publish methodology, not a universal savings percentage.
About
We build runtime control and security for AI agents.
REA Research Labs, Inc. was founded in July 2026. The hard problems in running AI agents are infrastructure problems, not model problems: runaway loops and retries, calls that should never leave, and agent activity no one can audit. REA ControlPlane handles them between the agent and the model.
Mission
Give every team running AI agents runtime control and security: cap spend, enforce policy before the call, and audit calls that use tools, at the layer between the agent and the model.
How we work
The right layer
We operate on structure — the AST and the token path — not on prompt wording. That is why two loops with different comments can still match.
Accountability
REA Research Labs, Inc. is a registered corporation. We stand behind the key issuance process and what the public API actually does today.
Who builds it
A small founding team that works on compilers and code structure, distributed caching, and LLM inference. Our starting point: loop traps are a problem of code structure, not of wording. That still drives the design, and it is why PreFlight and TraceGuard decide on your declared rules and on categories, not on message text.
We hire slowly, and only for hands-on work in inference systems, caching, or compiler tooling.
Company
| Legal name | REA Research Labs, Inc. |
| Founder & CEO | Elijah Godsil |
| Founded | July 2026 |
| Flagship | REA ControlPlane |
| Category | Runtime control and security for AI agents |
| Controls | LoopGuard, PreFlight, TraceGuard, included in every tier |
| Site | https://reacontrolplane.com |
| API | https://api.reacontrolplane.com |
Contact
Access, keys, and enterprise: elijah@reacontrolplane.com