AI Provider Keys

Common API key and AI provider problems and how to resolve them.

"No API key configured for provider"

Symptom: AI features fail with a message about no key being configured.

Cause: The active profile does not have an API key linked, and no environment variable fallback exists.

Solution:

  1. Go to Settings → AI Routing Settings.
  2. Select the provider you want to use.
  3. Enter your API key and save.
  4. Make sure the profile is set as active (default).

"API key rejected by provider"

Symptom: Requests fail immediately with an authentication error.

Causes:

  • The API key is invalid, expired, or revoked
  • Extra whitespace was pasted with the key
  • The key was saved under the wrong provider
  • A custom endpoint URL is incorrect

Solutions:

  1. In Settings, show the key to verify it looks correct (no extra spaces).
  2. Confirm you are pasting the key for the right provider (OpenAI key into OpenAI, not Anthropic).
  3. Test the key in the provider's own dashboard or playground.
  4. If using a custom endpoint override, clear it to rule out URL issues.
  5. Re-paste the key from your provider account.

"Rate limited by provider"

Symptom: Requests fail with a rate limit message, especially during agent execution.

Cause: You have exceeded the provider's rate limits (requests per minute or tokens per minute). This is common with free-tier keys or when running agents with multiple AI steps in quick succession.

Solutions:

  • Wait a moment and retry
  • Reduce max tokens in your profile settings
  • Upgrade to a higher tier with your provider
  • For agents with many AI steps, consider using autonomous mode (which naturally spaces requests)

"Provider billing issue"

Symptom: Requests fail with messages about insufficient credits or billing.

Cause: Your provider account has run out of credits or has a billing problem.

Solution: Check your provider dashboard for billing status. Add credits or update your payment method, then retry.

Works in chat but not in agents (or vice versa)

Symptom: AI works in one part of the app but fails in another.

Cause: ThinkForge supports per-interface routing — chat, agent creation, and agent execution can each use different provider profiles. If one interface has no key configured while another does, you will see this behavior.

Solution: Check the active profile for each interface in Settings → AI Routing Settings. Make sure each interface that needs AI has a valid key assigned.

Model list empty or refresh fails

Symptom: The model dropdown shows no options, or "Failed to refresh models" appears.

Cause: The key is invalid, the provider is unreachable, or a custom endpoint does not support model listing.

Solutions:

  • Verify the API key is valid and saved
  • Check your internet connection
  • Clear any custom endpoint override if you are not using one intentionally
  • The default model will be used as a fallback until the list can be refreshed

"AI request timed out"

Symptom: Requests time out after a long wait.

Causes:

  • Very high max token limit producing extremely long responses
  • Provider experiencing slowness or outage
  • Network connectivity issues

Solutions:

  • Reduce max tokens in your profile settings
  • Check the provider's status page for outages
  • Retry after a moment
Key after rotation: If you rotated your API key at the provider, remember to update it in ThinkForge. The old key in Settings will continue to be used (and fail) until you replace it.

Related