Problem The biggest hurdle for long-horizon agents is that they keep hitting setup walls for external APIs and services. Every new API requires account creation, subscription management, and API key provisioning — all designed for humans clicking through web forms. Agents can't open accounts for themselves, can't manage subscriptions, and can't provision their own credentials.

Today's auth systems assume a human is present: OAuth flows pop up browser windows, API dashboards require manual key rotation, and billing is scattered across dozens of individual provider accounts. None of this works when your agent needs to autonomously call 10 different APIs over a 6-hour task.

We need an agent-era auth system — one where a human sets up identity and payment once, and agents can immediately access any compatible service without per-provider onboarding.

Goals Agent self-registration — Agents can register for new services on their own, without human intervention for each provider Agent auth — A standard authentication mechanism that agents can use autonomously across any compatible service Payment (fiat first) — Built-in billing and settlement using fiat currency, so agents can consume paid services without per-provider payment setup Security of agent access to services and payment — Humans retain control over what agents can access and spend, with revocation, budgets, and audit trails Non-Goals Per-transaction settlement — The protocol settles in batches (weekly/monthly), not on every individual request. Real-time per-transaction payment adds unnecessary complexity and latency.