Salus · Runtime authorization for AI agents — the layer that decides whether an agent's *next action* is allowed to touch a real system, which is the control problem every team hits the week after their agent starts writing.

Sits between an AI agent and the system it is about to change. When the agent proposes an action — issue this refund, book this flight — Salus checks it against policy and against the evidence behind it, then blocks it, hands the agent the exact missing step, or lets it through, and writes a receipt either way. You wrap the function that performs the write; your model, agent loop and backend stay as they are.

Sector

Security & governance

Stage

Seed

On the roster since

15 Sept 2026