Manifold is a runtime security platform designed to provide visibility and protection for AI agents operating on enterprise endpoints. As AI agents (such as coding assistants and autonomous workflow tools) become more prevalent on employee devices, Manifold addresses the security gap created by their non-deterministic, autonomous behavior. The platform focuses on monitoring and securing these agents at runtime — meaning it observes agent behavior as it happens, rather than relying solely on pre-deployment screening or static analysis. Key aspects of what Manifold does: - Provides runtime visibility into the behavior of AI agents running on endpoints - Detects risky or unauthorized agent actions, such as unauthorized system access and improper data exposure - Operates agentlessly, meaning it does not require a traditional endpoint agent to be installed to function - Targets enterprise environments where employees may connect AI tools to internal systems without IT approval (Shadow AI) - Addresses the security challenges posed by autonomous, probabilistic AI systems that traditional endpoint security tools are not designed to handle The platform is positioned as a detection and response layer specifically for the agentic AI attack surface on endpoints, complementing existing security infrastructure rather than replacing it. It is aimed at security teams (CISOs, security engineers) in enterprises where AI agent adoption is outpacing governance and visibility controls.