AI Agents Chat: Error Messages & Troubleshooting

Errors in AI Agents chat can happen for a number of reasons — from a momentary blip with the underlying AI provider to a credential that has expired on a connected ad platform. Most errors resolve with a simple retry. This page walks through the error messages you may encounter, what they mean in plain terms, and what to do about them.

If an error doesn't match any of the categories below, or if you've tried the suggested steps and the issue persists, see When to Contact Support at the bottom of this page.

Rate Limit Errors

• What you'll see: A message indicating you've hit a rate limit.
• Why it happens: Your account or organization has sent more messages than the current rate limit allows. Limits reset on a rolling window.
• What to do: Wait a short time and retry the message. If you regularly hit rate limits, consider whether long, frequent prompts could be combined into fewer, more specific asks.

Provider Overloaded

• What you'll see: A message indicating the underlying AI provider is overloaded, with an inline link to that provider's live status page (Anthropic, OpenAI, or Google, depending on which model your Agent uses).
• Why it happens: The AI provider that powers your Agent is experiencing high demand or a temporary service degradation on their end.
• What to do: Wait a few minutes and try again. The linked provider status page will tell you whether the issue is widespread.

Timeout or Max Iterations Reached

• What you'll see: A message that the Agent timed out or reached its maximum number of reasoning steps. A Continue generating button may appear.
• Why it happens: The Agent ran longer than its processing limit, or used up its allotted reasoning steps before producing a final answer. This is most common with very broad or multi-part requests.
• What to do:
  • If a Continue generating button is shown, click it to let the Agent pick up where it left off.
  • Otherwise, simplify the request or break it into smaller, more focused questions.

Context Window Exceeded

• What you'll see: A message indicating the conversation or input is too large to process.
• Why it happens: Each AI model can only consider a limited amount of text at once (its "context window"). Long conversations or very large pasted-in content can exceed it.
• What to do:
  • Start a new conversation to reset the context window.
  • Avoid pasting large blocks of text in a single message — upload a file or break content into smaller chunks instead.
  • See Chatting With Agents for more on context windows.

Tool or Data Failures

• What you'll see: A message indicating the Agent tried to query a dataset or use a tool and it failed.
• Why it happens: The connected dataset is unavailable, returned an error, or doesn't contain the data the Agent expected.
• What to do:
  • Confirm the dataset connected to the Agent has data for the date range and accounts you're asking about.
  • Verify the dataset is still active in NinjaCat.
  • Rephrase the question with more specific scope and try again.

Service Restart or Temporary Unavailability

• What you'll see: A message indicating the AI Agents service was briefly unavailable mid-conversation.
• Why it happens: The service restarted while your message was being processed.
• What to do: Refresh the page and retry the message. Your conversation history should be preserved.

Network or Connection Errors

• What you'll see: A message indicating a network or connection problem.
• Why it happens: A client-side network interruption or browser connectivity issue — not something on the NinjaCat side.
• What to do:
  • Check your internet connection.
  • Try an incognito or private browser window to rule out browser extensions.
  • Clear your browser cache and cookies for ninjacat.io.
  • Try a different browser (Chrome, Firefox, Safari, Edge).

Authentication or Session Errors

• What you'll see: A message indicating your session has expired or you need to authenticate again.
• Why it happens: Your NinjaCat session has timed out or an authentication check failed.
• What to do: Log out and log back in. If the error continues to appear after a fresh login, contact NinjaCat Support.

Credential or Integration Failures

• What you'll see: A message indicating a connected ad platform credential is invalid, expired, or revoked.
• Why it happens: The credential the Agent relies on to query a data source is no longer valid — for example, a token has expired or access was revoked at the platform.
• What to do: Go to the connected data source in NinjaCat and re-authenticate the credential. Once the credential is restored, retry the Agent.

File Upload Errors

• What you'll see: A message indicating a file couldn't be uploaded — most often because it's too large or in an unsupported format.
• Why it happens: Files in chat must be under 30 MB, and only supported file types can be uploaded.
• What to do:
  • Reduce the file size to under 30 MB (compress, split, or export a smaller subset).
  • Confirm the file type is supported — see Chatting With Agents for guidance on file uploads.

"Something went wrong" — Generic Fallback

• What you'll see: A generic "Something went wrong" message. Occasionally, you may see raw error text passed through from the underlying AI provider.
• Why it happens: This is a catch-all for unexpected failures that don't fall into one of the categories above.
• What to do:
  • Refresh the page and retry the message.
  • If the error keeps appearing across multiple retries, contact NinjaCat Support with the timestamp and a description of what you were asking.

Agent Management Errors

• What you'll see: A message indicating the Agent's configuration couldn't be loaded or initialized.
• Why it happens: Something went wrong while loading the Agent's setup — for example, a misconfigured Agent or a backend issue retrieving its settings.
• What to do: Refresh the page. If the Agent still appears broken, contact your NinjaCat admin so they can review the Agent's configuration in Agent Builder.

Trigger or Automation Errors

• What you'll see: A message indicating a scheduled or triggered Agent run encountered an error.
• Why it happens: A trigger fired but the Agent run failed — often due to a configuration issue or an inactive connected dataset.
• What to do:
  • Review the trigger configuration in the Agent Builder.
  • Verify the connected datasets are active and contain data.
  • Manually run the Agent once to confirm it can complete successfully outside of the trigger.
  • See Adding Triggers for more on configuring triggers.

When to Contact Support

If the steps above don't resolve your issue — or if multiple users in your organization are hitting the same error — open a support ticket with NinjaCat Support and include:

• Which Agent you were using.
• What you were asking the Agent to do (the prompt or a description).
• The exact error message text (a screenshot is ideal).
• A rough timestamp (date and approximate time, with timezone) so we can trace it in our logs.
• Whether the issue is reproducible or happened only once.
• Whether other users in your organization are seeing the same error.

The more of this information you can include up front, the faster Support can diagnose and resolve the issue.