Reco Raises $55M to Map Shadow Agents and Agent Permissions
Reco has raised $55 million in new financing to expand its work on AI agent security. The company announced the round on September 29, 2026. TechCrunch
The financing builds on a $30 million Series B raised in February. The extension drew investment from AT&T, a Reco customer investing through its venture arm, alongside Forestay and Quadrille. That customer-investor detail matters. It points to production deployment rather than pilot interest.
Reco is led by co-founder and CEO Ofer Klein. Under Klein, the company has repositioned from software to map and secure SaaS and AI platforms to a broader architecture centered on a context graph. The graph connects agents to applications, people, accounts and permissions. For practitioners, the model is familiar. It treats agents as first-class identities with inheritance, delegation and scope, not as isolated API clients.
Discovery is the hard part. Reco said its platform found 21,000 agents that a Fortune 100 customer did not know about. That number matters. Most enterprises already struggle with SaaS sprawl, OAuth grants, service accounts and over-permissioned integrations. Autonomous agents multiply the problem because they can be created by business users, developers and other agents, often outside centralized IT provisioning.
The February Series B was led by Zeev Ventures, with participation from Insight Partners and boldstart ventures. That syndicate was disclosed in April. The September extension adds strategic and cross-stage capital without replacing the earlier lead.
The current round has history. In April 2025, Reco described an additional $25 million raise and said it intended to use the funding to scale its AI-native SaaS offering and to close what it called the SaaS Security Gap with Dynamic SaaS Security. Reco Redseed participated as a new investor in that $25 million round at the time. Calcalistech
Operational signals since then include recognition as a CRN 2025 Stellar Startup and a doubling of headcount across multiple regions. The company has also been named an OpenAI Select Partner, lists a catalog of more than 270 agents and apps, and says its Factory provides more than 260 app integrations. Those catalog and integration counts are relevant to coverage. In SaaS security, efficacy tracks closely with the breadth and depth of API connectors and the fidelity of permission parsing.
The broader context here is a crowded field. Startups are racing to define agent security, with approaches ranging from runtime guardrails and prompt filtering to identity governance and posture management. Reco is betting that the control point is the relationship map. If you can continuously resolve which agent can act on which data, on whose behalf, and through which app authorization, you can enforce policy before execution and audit after it.
In my view, that emphasis is sound for an expert buyer. Inference controls and content filters have a role, but enterprise risk from agents will concentrate in authorization logic. Stale OAuth tokens, excessive scopes, shared service accounts and transitive access through connected apps are already known failure modes. Agents make them dynamic. A context graph is one plausible way to keep the permission model current as agents spawn, share credentials and chain tools. The open questions are operational. How fresh is the graph, how noisy are the findings, and how cleanly can remediation plug into existing identity and ticketing workflows without creating a parallel console that security teams ignore.
