OAuth was built for a closed-world deployment model where clients, authorization servers, and resource servers mostly knew each other before runtime. Agents are pushing OAuth toward an open-world model, and that evolution brings two challenges: the protocol substrate and the Mission governance layer above it.
OAuth was built for closed worlds, and that constraint is why it became mature. Agents expose the limits of that deployment model. This post traces what the newer OAuth standards get right and which substrate gaps still need to close.
Open-world OAuth can improve discovery, resource binding, and first-contact trust. That still leaves the harder agent problem: how approved intent becomes bounded authority that stays governed across delegation chains, unfamiliar tools, consent expansion, revocation, and task termination.