Caura provides governed memory for fleets of AI agents, enabling knowledge learned by one agent to be shared with authorized agents across an organization. Designed to work with different models, agent frameworks, and providers, it offers persistent memory through MCP, REST, and developer tools, with a free tier and an Apache 2.0 open-source core that organizations can run themselves.

Governance is built into memory operations. Caura enforces tenant isolation, fleet-level access policies, agent identity and permissions, and audit logging for reads, writes, and deletions. Content policies can detect and flag, mask, or drop sensitive information, while tools surface contradictions, stale facts, and provenance.

Caura Prism is the enterprise console for visibility, audit, risk, and spending. It helps teams inspect what agents know, trace recalls, review security findings, and understand costs. The platform also offers developer documentation, an MCP server, SDK integrations, and enterprise deployment options, including on-premises, air-gapped, and white-label configurations with dedicated support.