SpiderGate now retires a failing LLM key automatically and re-authenticates it in place
A dead provider key is withdrawn from the shared routing pool without human intervention, and the contributor who supplied it can restore it without rebuilding the entry
WILMINGTON, Del., June 26, 2026
TALLINN, Estonia, June 26, 2026 — SpiderGate, the OpenAI-compatible LLM gateway operated by DEMAVIAS, LLC, today detailed the credential-health layer that removes a failing provider key from its routing pool without human intervention, distinguishes a temporary provider fault from a revoked credential, and lets the person who supplied the key restore it in place rather than rebuilding it.
The behaviour addresses a failure that is difficult to observe from the outside. When a provider API key stops authenticating inside a shared pool, requests do not fail; they fall through to the next model in the chain. Work continues on a different model, at a different cost, with no outage to alert anyone.
"A dead key is not an outage, which is exactly what makes it expensive," said Martin Shein, founder of SpiderIQ. "Nothing goes red. Your agents keep answering, slightly worse, on a model you did not choose, until somebody happens to open the vault. I wanted the pool to notice before the customer does."
Three consecutive failures mark a credential unhealthy and routing skips it. A credential that has been unhealthy for twenty minutes with no authentication errors returns to the pool automatically, on the basis that rate limits and timeouts resolve on their own. A credential that fails authentication three times within twenty-four hours is deactivated outright and cannot be automatically restored, which ends the recover-and-fail cycle that a revoked key would otherwise repeat roughly hourly [source: https://gate.spideriq.ai/docs/reference/key-health].
Notification is directed rather than broadcast. The contributor who supplied the key receives a re-authentication link at most three times, 48 hours apart, and never inside their configured quiet hours [source: https://gate.spideriq.ai/docs/reference/key-health]. Brand administrators instead receive a single daily digest at 09:00 in their own timezone, listing every unhealthy key alongside the contributor responsible for it [source: https://gate.spideriq.ai/docs/dashboard/key-health].
Re-authentication updates the existing credential rather than replacing it. The key identifier, sharing settings, limits, usage policy and history are preserved, and only the secret changes; the routing engine reloads without a restart. The emailed link remains valid for 120 hours, chosen to outlast the full three-reminder cycle, and an expired link can be reissued to the contributor on record without administrator involvement [source: https://gate.spideriq.ai/docs/reference/key-health].
"The part I would defend hardest is who gets the email," said Martin Shein, founder of SpiderIQ. "We sent admins one message per failing key once, and one person received fifty in a day. An alert that arrives fifty times a day is not an alert, it is a filter rule. Now the link goes to whoever can actually replace the key, and everyone else gets a summary."
The routing engine underneath SpiderGate is BerriAI's litellm.Router; the credential-health layer, task-alias routing, per-tenant authentication and usage records are SpiderGate's own.
About SpiderGate
SpiderGate is the LLM gateway built for agencies: provider keys live encrypted in a Vault that contributors can add to without exposing them, and a shared pool with least-busy failover keeps requests flowing when a key rate-limits or dies. SpiderGate is operated by DEMAVIAS, LLC.
Media contact
Press Office, Media Relations Email: press@spideriq.ai Web: https://gate.spideriq.ai/press
About SpiderGate
SpiderGate is the LLM gateway built for agencies: keys live encrypted in a Vault you contribute to without exposing them, and a shared pool with least-busy failover keeps requests flowing when a key rate-limits or dies. On top sits Studio — a chat workbench where you switch between 1,200+ models mid-conversation, with your files, notes, and saved prompts organized by project.
Media contacts
For interviews, assets or further information.