Reco extends AI agent governance to Claude across enterprise stacks

Reco, a Florida-based AI governance platform, has launched a bidirectional security integration with Anthropic's Claude, connecting activity across Claude Enterprise and Claude Platform to the identities, permissions and applications that determine enterprise risk. The move reflects a broader structural shift: agentic AI is no longer a productivity add-on sitting at the edge of corporate infrastructure. It is becoming, as Reco's own CEO puts it, part of the enterprise operating fabric.

The integration works across two administrative surfaces. Claude Enterprise is where employees interact with Claude day to day; Claude Platform is where developers manage API keys, workspaces and agent deployments. Reco ties signals from both into its Reco Graph, a real-time map of identities, permissions, application connections and activity spanning more than 230 enterprise tools. The aim is to catch risks that single-source monitoring misses: an overpermissioned agent with a path to sensitive data, an API key that outlives its original deployment, or a former employee whose Claude access was never revoked after offboarding.

The security gap that agentic AI opens

The announcement surfaces a governance challenge that has accelerated faster than enterprise security teams anticipated. When employees used Claude as a conversational interface, risk was relatively bounded. The expansion into agentic deployments, where Claude-powered agents take actions across connected systems via Model Context Protocol (MCP) servers, changes the blast radius of any single misconfiguration.

Reco addresses six risk categories it says are driving enterprise security conversations: shadow AI and unauthorised usage, sensitive data leakage, prompt injection, API key exposure, excessive agent permissions, and access-control gaps around offboarding. Its integration with Claude's Compliance API gives security teams structured visibility into Claude Enterprise activity, while the Claude Platform integration exposes the development layer where agent configurations are set and API keys are issued. Reco also maps each agent's model version, tool access, permission policies and MCP server connections, then correlates that context across the wider application estate.

A notable capability is the natural-language investigation interface: Reco's MCP server allows security teams to query the Reco Graph through Claude itself, asking plain-language questions about anomalous access or risky agent connections and routing findings into existing SIEM, SOAR and ticketing workflows. "Security teams need to understand who is using it, what agents and applications it connects to, what permissions are involved, and what risk is created when AI activity moves across the business," said Ofer Klein, CEO and co-founder of Reco.

Convergence read-across: cybersecurity meets the agentic enterprise

The broader significance for cross-sector strategists is not the product itself but the category it confirms. Enterprise AI governance is rapidly consolidating into a distinct security discipline, adjacent to but distinct from traditional identity and access management (IAM) and SaaS security posture management (SSPM). Reco positions Claude alongside Okta, Salesforce and Microsoft 365 as a critical enterprise application requiring the same governance rigour, a framing that signals where the security-software market is heading as foundation-model vendors deepen their enterprise integrations.

For investors tracking the cybersecurity and AI infrastructure intersection, the competitive landscape is filling quickly. The agentic-AI security category is attracting dedicated funding as enterprises accelerate deployment of autonomous workflows and as foundation-model providers, Anthropic included, build out compliance and audit APIs specifically to support third-party governance tooling. Regulatory pressure is a tailwind: the EU AI Act's obligations around high-risk automated systems and audit trails are pushing procurement teams to formalise AI governance before regulators require it. That dynamic is likely to compress the window for enterprises still treating agentic deployments as informal pilots. Capital flowing into this space is following the same convergence logic that has already reshaped cloud security: the attack surface is defined by connectivity, and governance tools that map the full graph of identities, permissions and agent actions will command enterprise budgets at a scale that point solutions cannot.

The Reco Platform with Claude Security integration is available immediately through Reco and its channel partners globally.