Bay in context: an endpoint-only AI control delivered through EDR

What do Bay's pages describe?

Bay calls itself a fully preventive, agentless platform for AI agents and tools on enterprise endpoints. Its pages describe three layers: an inventory of AI agents, MCP servers, tools, credentials, extensions, settings and endpoint deployments; activity monitoring that records who triggered an action, whether a human was involved, and the chain from prompt to system action; and enforcement that returns Allow, Ask or Deny.

The enforcement detail comes from Bay's August 2026 research post. It says its Enforcer evaluates prompts, pre-tool calls and post-tool responses and returns Allow, Ask or Deny locally, which it states takes under 4ms, on supported agent surfaces. Capability rules cover shell and code execution, process spawning, package installation, cloud CLIs, containers, Kubernetes, browser automation and system changes. For Claude Code, Codex and Claude Desktop, administrators can lock permission rules, preserve managed hooks, restrict MCP servers and constrain plugin sources. New rules can run in Simulation Mode.

Delivery is the other distinctive part. Bay says it deploys through your existing EDR or MDM, with no new agents and no new tools, as an ephemeral binary it calls Wave that scans and exits.

Where does that fit?

In the finder, Bay matches the device position fully and every device-related check it can: naming coding agents, tool-call decisions, rollout through existing EDR or device management, ask, block, a watch-only or simulation mode, and an action record. In our example profiles, it comes first alone for a company whose coding agents can reach production, which wants to block first and runs EDR and device management. It comes second in the Palo Alto Networks and SentinelOne profiles, behind the product built into that platform. That reflects how the rules treat a product whose pages describe decisions on the device in detail.

Where does it not fit?

Anywhere the risk sits outside the device. Bay's pages describe no browser extension, no network proxy and no gateway, so a company whose main concern is employees pasting data into browser AI, or whose work runs in cloud workspaces, will find products that describe those positions higher in its results. In our default example, which combines browser AI and coding agents with a preference for coaching, Bay is not in the first group. It also describes nothing for SaaS agents, cloud-hosted AI apps or homegrown agents, where Noma Security, Zenity, Onyx Security and Lasso Security describe coverage.

What are the limits to weigh?

• Endpoint only, as above.
• One of the smallest public footprints here: a single homepage, two blog posts, no documentation portal and no published pricing. Ask for architecture documents and references during evaluation.
• Headline figures such as under 5 minutes to full deployment and a false positive rate under 1% are Bay's own statements, and we have not verified them.

How does it compare with the other EDR-delivered product?

Noma Security also states that it works through existing EDR or MDM without a new endpoint agent. The difference on our checks is in decisions and breadth: Bay's pages describe ask, a simulation mode and a person-versus-agent record, which Noma's pages do not describe; Noma's pages describe SaaS and homegrown agents, which Bay's do not. The Bay and Noma head-to-head shows every recorded fact side by side.

Does Bay replace EDR?

No. Bay's pages describe deploying through EDR or MDM, and its decisions are about what AI agents may do, not about malware or intrusion. Plan for both.

What should you ask Bay in an evaluation?

1. Which agent surfaces does the Enforcer support today, and which are next?
2. What exactly runs on the device, for how long, and what does it change?
3. What does a developer see when an action is set to Ask, and how fast is the answer recorded?
4. Can you share architecture documentation?
5. What would cover the browser, cloud workspaces or SaaS agents alongside Bay, if we need them?

These are the same kinds of questions we suggest for every product, sharpened by where Bay's public pages are thinnest.