Reco Raises $55M to Secure the AI Agents Enterprises Can't See

Reco, a startup that helps companies find and secure the AI agents running inside their networks, raised $55 million in new funding on September 29, 2026. The round was led by Adams Street Partners, with a strategic investment from AT&T Ventures and participation from new investors Forestay and Quadrille Capital, according to BigGo Finance. The raise brings Reco's total funding to $140 million.

The company says the money will go toward expanding sales, partnerships, distribution channels and customer support. The bet behind the round is simple: enterprises are deploying AI agents far faster than their security teams can track, and the market for watching those agents is growing quickly.

What Reco does

Reco's platform identifies the AI agents operating inside a company, maps their identities and permissions, and shows what data and systems each agent can reach. It then visualizes business workflows so security teams can see where risk concentrates. The platform connects to more than 280 applications, including OpenAI, Anthropic, Microsoft Copilot, Salesforce, ServiceNow and Workday, and ships with 1,000 detection controls, BigGo Finance reports.

The core technology is called the Reco Graph. It maps the relationships between agents, identities, applications, permissions, data and workflows as a connected graph. The company's pitch is that this lets security teams identify and cut only the risky connections — an agent with access it should not have — without breaking the legitimate automation the business depends on.

Two companion products round out the offering. Reco Factory lets new application integrations be added within hours, and the Reco Library is a pre-populated repository of information about agents and applications that gives security teams a starting inventory instead of a blank page.

The 21,000 agents nobody knew about

The most striking detail in the funding coverage comes from Reco co-founder and CEO Ofer Klein. He told AI Daily Post that at one Fortune 100 customer, Reco's platform found 21,000 agents the company did not know it had. That number captures the core problem: agents are being built and deployed by individual teams, vendors and employees faster than any central inventory can follow.

Klein points to that shift — companies deploying agents faster than they can track them — as the reason Reco broadened its scope. The startup spent its early years mapping SaaS and AI platforms for enterprise security teams. Now it is rebuilding its pitch around a larger idea: a context graph linking agents to the apps, people, accounts and permissions they touch, so security teams can see exactly what each agent can reach and shut off access it does not need.

That pivot mirrors a wider industry turn. Agent capability has outrun agent oversight, and the companies selling oversight are multiplying.

A crowded field with the same diagnosis

Reco is entering an increasingly crowded market. Crunchbase and PitchBook list at least two dozen startups selling some form of AI agent security, from tool vetting to data access controls to runtime detection, with established security names like CrowdStrike building their own offerings, AI Daily Post notes. The pitches overlap heavily: knowledge graphs, continuous monitoring, governance dashboards.

The overlap is itself a signal. For chief information security officers already managing the sprawl of agents across their networks, the growing list of vendors promising to tame that sprawl is becoming its own kind of sprawl. Differentiation will come down to which platforms can show concrete results — agents found, permissions trimmed, incidents prevented — rather than architecture diagrams.

Reco's answer is specificity: find the agents first, map what they touch, and remediate the risky links without disrupting the rest. Whether that focus is enough to stand out in a field of two dozen will be the story to watch.

Security is becoming the agent economy's bottleneck

The Reco raise lands in the middle of a busy week for agent infrastructure. Enterprises are moving agents into production in regulated work — Cognizant just took agentic claims processing live for health plans with audit trails built in, and Mastercard's Agent Pay program is putting programmable consent onto live card rails. Each of these moves assumes the same thing: agents earn autonomy in proportion to the oversight around them.

Reco is attacking the problem from the other direction. Instead of building safer agents, it watches the agents companies already have — including the 21,000 nobody knew about. As agent deployments keep compounding, the inventory problem only gets harder, and the companies that can see the whole map will set the terms for everyone else.