Reco Raises $55 Million as AI Agent Security Becomes a Crowded — and Urgent — Market

Reco, a startup that helps companies find and control the AI agents multiplying inside their systems, has raised $55 million — an extension that builds on the $30 million round it announced in February and lifts total funding to $140 million. AT&T Ventures, whose parent is also a customer, invested alongside Forestay and Quadrille Capital. Co-founder and CEO Ofer Klein said Reco's valuation has "more than doubled" since February into the high hundreds of millions, without giving exact figures.

The pitch responds to a problem security teams describe with growing alarm: enterprises are building and deploying AI agents faster than they can inventory them. Klein says Reco's platform, built around a context graph linking agents to apps, people, accounts and permissions, once surfaced 21,000 agents at a Fortune 100 customer that the company had no record of. At a large financial services client, Reco says it identified an agent created by a former employee that could still reach Salesforce and exfiltrate data to a domain the company couldn't see. (The customer examples are company-reported and not independently verified.)

Reco's product has shifted accordingly. Until last year it mostly mapped and secured SaaS platforms; now it positions itself as end-to-end agent governance — using browser and network signals to find agents operating outside the 280-plus apps it integrates with, and offering controls to inspect prompts and tool calls, and to revoke access an agent doesn't need. Klein says annual recurring revenue is in the "double-digit millions" and expected to triple this year, with more than 100 customers, about 40% in financial services (company figures).

Reco is far from alone. A quick scan of Crunchbase and PitchBook turns up at least two dozen vendors selling some form of agent security, with nearly identical vocabulary — knowledge graphs, continuous monitoring, runtime security, tool access controls, MCP vetting. CrowdStrike is building detection and response on the devices agents run; HiddenLayer's CEO says over 50 of its customers have agents in production touching critical systems; Cymphony says it found some 85,000 files newly accessible to AI tools at one US public company.

The crowding itself is the signal. Six months ago the question facing a CISO was whether to buy agent security at all; now it's which of a dozen vendors to pick, at exactly the moment agent deployments — and agent incidents, from OpenAI's Australian government intrusions to Meta's Muse privacy lapses — have made the category feel less optional. Vendors promising agent discovery are, in effect, selling the map of a territory most enterprises haven't surveyed.

Reco's wager is that its SaaS-native roots give it a head start: it already has coverage of the apps where agents live, and claims new integrations can be added within days. The $55 million buys runway to prove that before the platform giants fold this capability into existing security bundles — the usual way crowded categories consolidate. For now, the money and the anecdote land the same message: nobody, including the enterprises deploying them, knows how many AI agents are inside corporate walls.