Keep serving.
A deterministic gateway checks health and switches to a preapproved standby. It works while the agent, discovery network, or payment rail is unavailable.
Health → switch → serve
NeverDown / programmable resilience
When a service fails, your app keeps going.
Define what NeverDown may do before a dependency breaks. It switches to an approved standby first, then finds and verifies replacement capacity under your policy.
NeverDown / recovery
Example flow
One failure. A controlled response.
-
×
Primary becomes unhealthyHealth checks detect a failed dependency.
-
✓
Approved standby takes overRequests keep flowing through the fast plane.
-
↗
Capacity is replenishedVerify an offer, check policy, pay, and install.
Recovery does not wait for a marketplace or payment.
One policy.
Two independent clocks.
Restore redundancy.
The Supreme Agent looks for a replacement, verifies the seller and capability, and asks the Resilience Contract to authorize any purchase and installation.
Discover → verify → authorize → installEvery step has a reason.
Watch the spot
Measure the health signals that matter for the dependency.
Switch safely
Route to a healthy standby already approved by the operator.
Prove before paying
Check offer, identity, and live capability before a capped purchase.
Keep the receipt
Record policy decisions, settlement, and installation evidence.
Beyond one dependency
Start with RPC. Build for every critical spot.
NeverDown's first working adapter handles JSON-RPC. The same policy and recovery model can extend to other services as their adapters are built and verified.
RPC gatewayDemo adapter
IndexerFuture adapter
Price APIFuture adapter
The policy that Never let you Down
Open console ↗The hosted console is a preview. Live operator actions run in the local backend.