Archestra Raises $10M Seed for AI Agent Data-Access Guardrails

Archestra announced a $10 million seed round led by 20VC, bringing total funding to $13.5 million. The company describes its open-source platform as a horizontal layer for connecting AI agents to corporate data with guardrails, governance, and observability.

The product sits between an agent and internal systems. When an agent requests data, Archestra checks identity and policy, brokers access, and logs the result. The company says multiple Fortune 500 companies are already using the platform in production, with early traction in legal, procurement, operations, and regulated finance.

The DLP Angle

This is not traditional endpoint DLP, and Archestra is not positioning itself as a legacy DLP replacement. The relevance for data-protection teams is the control point: AI agents need access to SaaS, documents, tickets, source repositories, and other sensitive systems, but giving agents broad credentials creates obvious leakage and prompt-injection risk.

Archestra’s bet is that agent access needs a broker that understands policy, auditability, and deterministic guardrails before data reaches the model or tool chain. For DLP teams, that raises a familiar question in a new location: should enforcement happen at the endpoint, the browser, the SaaS API, the identity layer, or the agent gateway? In practice, mature programs will likely need several of those controls to agree on what data an agent can see and where it can send it.