Barrikade

Barrikade is building a security layer for AI agents. Its stated aim is to find AI agents within an organization, connect each one to an owner, detect unsafe behaviour, and record exactly what happened. Products and status: - Barrikade Lens: an open-source tool that discovers AI agents across endpoints, repositories, CI pipelines and Kubernetes. - Barrikade Core: a runtime security engine currently in working-alpha. It began as prompt-injection detection and has grown into a system that can monitor a full agent session, identify when an agent moves off task, and record the response. - Identity layer: planned and still under development. It is intended to link agents to owners. Background and direction: Barrikade Core started as an undergraduate research project, developed with design input from Jentic. The company is now working to connect Lens, Core and the identity layer into a single lifecycle covering discovery, ownership, runtime monitoring and audit, so that these stages can share context. The company states that it keeps the maturity boundary of each component clear. It offers a waitlist and publishes a security model document on AI agent security.