Bay, a Tel Aviv startup, has come out of stealth (operating without a public launch) with $50M to build software that checks what an AI agent is about to do before it does it.
The company is less than a year old. The $50M total includes a $37M Series A, an early-stage venture round, led by Battery Ventures. Bay says it already has dozens of enterprise customers and names Micron, Peloton, ClickUp and Wix. Those numbers and names come from the company itself.
The product sits between an agent's decision and its execution. An agent is software that does more than answer questions: it calls tools, sends messages and changes records. That is why vetting actions differs from filtering a chatbot's text. A bad answer can be ignored, while a bad action may already have happened by the time anyone reads the log.
Bay has not said how deeply those firms use the product. Four recognizable names could mean company-wide deployment or a single team's trial, and the difference matters when judging a young company's traction.
The design has a familiar shape. Payment networks check a transaction before approving it, not after the money moves, and Bay is betting agents need the same gate. The hard part is calibration. A checker that is too strict leaves agents too hobbled to be worth deploying, and one that is too loose is a rubber stamp.
There is also a market question. If the platforms where companies build agents add their own approval layers, a standalone vetting vendor has to prove it can do the job better, or across more systems, than the defaults.