What is AgentSafe?

AgentSafe is an execution-authority gateway that sits between AI agents or APIs and their target services. It intercepts consequential HTTP requests (writes, payments, deletions), captures them as canonical intents, and asks the Decionis control plane — an independent execution authority — for a decision: ALLOW (forward once on a single-use grant), ESCALATE (hold for human approval via Presence), or BLOCK (refuse with a signed dossier). The gateway runs in three modes: a local demo authority for testing, shadow mode to observe Decionis decisions on live traffic, and enforcement mode where nothing executes without a verified grant. It installs via Docker, Helm, Homebrew, Linux packages, or standalone binaries. The project is Apache-2.0 licensed and includes a comprehensive test suite (the Compromised Principal Test) that validates the boundary against adversarial attempts. AgentSafe does not make policy decisions itself; it enforces decisions from Decionis, which operates the policy control plane and Presence human-verification layer as separate services.

Should you self-host AgentSafe?

AgentSafe scores 53/100 on selfhost.directory (how scoring works) and ranks #18 of 39 in API Gateway. It has 672 GitHub stars.

It shipped 9 stable releases in the past year. The latest, v0.5.0, came out on 27 October 2026. New versions land roughly every 2 days.

Setup is moderate: expect some configuration beyond a single command. An official Docker image is available. It is Apache-2.0-licensed and written mostly in Typescript.

Its closest alternative on selfhost.directory is Kong Gateway (82/100, 44 314 stars). It scores higher overall, so compare the two before you pick. See AgentSafe vs Kong Gateway side by side.