Authorizing an agent once, at sign-in, is not enough. An agent runs for a long time, calls many tools, and can be copied or revoked while it works. Access granted at the start of a session says nothing about whether the agent is still trusted on its hundredth call. Agent identity needs authorization that happens continuously, on every action.
The checkpoint in front of your tools
The Gateway MCP sits in front of the company's tools. On each call it checks the identity token and the single-use operation proof, refuses unknown tools and arguments outside a tool's allow-list, and refuses unconfigured tools when its default-deny setting is on. A gateway proxy can front MCP servers written in any language, so the checkpoint is not limited to one runtime.
A signed decision for every operation
For each enforced call the Gateway MCP asks the identity service for a signed, single-use boundary decision. The decision checks that the agent is still active, that its principal is still active, and that the operation is within its permitted services and clearance. If no valid decision comes back, the call does not go through. The identity service does not reuse a decision across calls, so a change in status takes effect on the very next operation rather than at the next login.
Revocation that lands quickly
Revoke an agent and its next token request and next boundary decision are refused. Signed revocation snapshots are refreshed faster for agents with a more dangerous risk tier. There is also a kill switch with a way back: suspend an agent, then reinstate it after its host is checked again, with reduced access first. Instead of the token, a service receives a gateway assertion, a single-use statement for one request made for that one service.
How Oathera helps
Oathera gives every AI agent an identity bound to the machine it runs on, approved by a named person, and checked every time it acts. Continuous authorization is the "checked every time it acts" promise made literal at the Gateway MCP and the identity service. Every call writes one audit event into a hash-chained log that is sealed, verified and exported exactly once to a SIEM or write-once store. Oathera works beside your identity provider and AI governance as the identity layer under them.
See it, or build it with us
Watch a stolen token fail, a replayed proof get refused and a tampered call get caught, in the simulator at sim.oathera.org. Our first design partners start in early November 2026. If you are putting agents into production, become a design partner.
Every AI agent, sworn in.