As AI agents start acting on their own, they need a way to prove who they are to the services they call, without reusing a human master API key or pretending to be a browser going through OAuth and CAPTCHA. LIME is built for exactly that: an identity and authentication layer designed for AI agents rather than people.
Accounts, agents, and passports
LIME does not throw away accounts, it layers identity. A person or organization holds an owner account (a user_id). Under it, each agent is given a stable, public agent identity (an agent_id) with its own public profile and a long-lived Agent Token. When an agent needs to reach a service, it authenticates to LIME with that opaque Agent Token and receives a short-lived, signed passport, a JWT whose sub claim is the agent_id.
How services verify a passport
Passports are signed with RS256, and services verify them locally against the public LIME JWKS, with no network round-trip to LIME on the hot verification path. Two design choices keep this tight:
- Domain binding: a passport is scoped to specific hostnames, so a token minted for one service cannot be replayed elsewhere.
- Short lifetime: passports expire quickly (on the order of ~300 seconds), limiting the damage if one leaks.
The practical pitch is that a platform can accept autonomous agents without handing out long-lived master keys or bolting agents onto human login flows, while still getting local, low-latency verification.
Who it is for, and pricing
LIME targets platform and SaaS developers who want AI agents to register and authenticate cleanly, and teams running fleets of agents that talk to many services. Per LIME, core agent identity, issuing passports and verifying them locally, is free forever; it notes that free covers identification, not necessarily every future product built on top. As with any young protocol, check the official site for the latest specifics before you build against it.
Comments
No comments yet
Be the first to comment