integrations

Troubleshoot an integration

Diagnose expired authorisation, provider errors, missing services and connections that show as connected but return no data.

· OpenbirdAI documentation

Almost every integration problem falls into one of four categories: the authorisation is invalid, the wrong account is connected, a required scope is missing, or the provider itself is unavailable. This guide isolates which one you have.

Work through these in order

  1. Check the connection status shown on the integration card.
  2. Confirm the connected account label is the account you expect.
  3. Confirm you are in the workspace that owns the connection.
  4. Reconnect if the status is expired, invalid or missing.
  5. Test with a single narrow question and inspect the sources list.
  6. If it still fails, check whether the provider is having an incident.

Connected but no data

  • The service exists but is not usable by the connected account — common for Analytics properties and Ads manager accounts.
  • The data exists but falls outside the requested date range.
  • The requested record is owned by, or shared with, a different account.
  • A required scope was declined at authorisation.

Provider errors

Providers return their own error classes. Rate limiting and quota errors are transient and worth retrying after a pause. Invalid-grant and permission errors are permanent until you reconnect. Authentication failures on the provider side usually mean the account lost access to the resource entirely.

Do not retry a failing authorisation in a tight loop. Repeated attempts can trigger provider-side rate limits and make recovery slower.

Still stuck

Record the provider, the approximate failure time, the exact error text, and whether a narrow test question failed too. That is the information support needs — and none of it should include tokens or passwords.

When the problem is the provider, not OpenbirdAI

Some failures have a well-understood external cause: a provider outage, a quota that exhausted for the whole account, a platform policy change, or an app review that lapsed. Check the provider status before escalating internally, because no internal change will fix it.

  1. Confirm the provider is reachable by testing an unrelated feature.
  2. Check whether your account-level provider quota is exhausted.
  3. Wait out a declared incident rather than retrying continuously.
  4. Escalate to support only once the provider side is healthy.

Common questions

The integration says connected but nothing works. What now?

Reconnect and approve all scopes, confirm the connected account label, and test a narrow question. If it still fails, the account likely lacks access to the underlying resource.

Is it safe to reconnect repeatedly?

Yes, but space attempts out. Rapid repeated authorisations can trigger provider rate limits.

Was this page helpful?