# Gateway operational readiness

Loop Gateway's current new-sales review remains in effect. Existing balances and obligations must be preserved. A reachable API or successful discovery check is not evidence of supplier permission, production readiness or paid fulfillment.

The improved delivery direction is a scoped, customer-owned operating deployment with tested boundaries, failure handling, recovery and an accountable maintenance handoff. It requires a documented supplier arrangement and confirmed capability before a production offer is accepted. No new managed Gateway subscription is being sold by this guide.

## Working discovery toolkit

Download Loop Evidence 1.1 and choose Observe sources → Gateway discovery checks. Enter the Gateway HTTPS origin. It records three read-only requests:

- `/health`: successful response with an ok/healthy status.
- `/v1/models`: a non-empty model catalog containing model IDs.
- `/v1/pricing`: a non-empty structured response; rate correctness is not verified.

Each observation retains its fingerprint, collection result and timing. Failed collection or a failed contract enters the exception queue. There are no top-ups, inference calls, wallet operations or funds transfers. Collection is on demand and establishes no uptime guarantee.

## Eight readiness criteria

1. Document supplier authorization and the proposed supply arrangement.
2. Agree customer ownership, access, responsibilities and service boundaries.
3. Pass the documented discovery contracts.
4. Test unauthorized use and exhausted-budget behavior.
5. Reconcile an explicitly authorized usage and settlement test if payments are in scope.
6. Test upstream errors, timeouts and uncertain-settlement handling.
7. Demonstrate recovery and rollback without exposing secrets.
8. Obtain operator acceptance of the runbook, maintenance boundaries and escalation procedure.

Loop Evidence requires a recorded pass with a reference for every criterion before its Gateway readiness record can be accepted. Automatic discovery results cannot satisfy the other criteria. No tests should spend money, disclose credentials or alter customer infrastructure without the applicable authorization.

Commercial viability must be measured against implementation labor, maintenance cost, dependency risk and an actual buyer's needs. Model count, a new interface, a margin on upstream tokens or a health endpoint alone is insufficient differentiation. No unsupported uptime SLA or supplier relationship is claimed.

Supplier reference: https://openrouter.ai/terms/