> ## Documentation Index > Fetch the complete documentation index at: https://aidocs.ethanbragdon.icu/llms.txt > Use this file to discover all available pages before exploring further. # Developer API Troubleshooting > Troubleshoot Starfire AI developer credentials, permissions, quotas, request failures, runs, webhooks, and API usage. # Developer API troubleshooting Use this guide when a Starfire developer integration fails, returns an unexpected status, or stops receiving asynchronous events. ## Start with the request identifier When the Developer Platform returns a request identifier, record it with the timestamp and endpoint family. That is the fastest way to connect your application's error with Starfire-side logs and diagnostics. ## Credential rejected Check: * the integration is using the intended credential * the credential has not expired or been revoked * the application or organization still owns it * the required permission scope is present * the owning account/organization still has Developer Platform access Do not paste the secret credential into support messages. ## Permission denied A valid credential can still lack permission for a resource or operation. Compare: * credential scopes * organization policy * resource ownership * plan entitlement * feature rollout Creating a broader credential should not be the first troubleshooting step. ## Rate or usage limit reached Check the applicable plan quota, credential/application override, concurrent-run limits, and credit state. If the limit is temporary, respect the retry/reset guidance returned by the active API contract. If a hard credit ceiling is reached, repeated retries will not solve it. ## Run/build accepted but later failed An accepted asynchronous request only confirms creation. Inspect the run/build/research resource for the final state and failure stage. ## Webhook not received Determine whether: 1. the Starfire event occurred, 2. a webhook delivery was attempted, 3. the endpoint responded, 4. Starfire scheduled a retry, 5. your consumer rejected or ignored a duplicate/event identifier. Use webhook delivery history where available rather than recreating the original operation. ## Unexpected duplicate event Webhook consumers should be idempotent because asynchronous delivery can retry the same logical event. ## High latency Break down whether latency comes from Starfire API processing, model/provider execution, long-running job state, or your own webhook/consumer infrastructure. ## Broad API failure If unrelated API clients and endpoints are failing at the same time, operators should check Developer Platform health and the incident system. Use request IDs and developer telemetry to investigate one integration. Use Control Center for platform-wide developer issues.