Skip to main content
Quick fixes for common implementation hurdles.
  1. “API Connection Timeout”: Usually a firewall/whitelisting issue. Ensure Atlas IPs are permitted in your warehouse.
  2. “Unknown Event/Field”: The event exists in your DB but hasn’t been “imported” into the Atlas Dictionary.
  3. “Empty Result Set”: Verify your timestamp format. Atlas expects ISO-8601 for time-series analysis.
  4. “Lumos Hallucination”: If the AI generates the wrong SQL, use the “See SQL” button to debug the join logic.
  5. “MCP Auth Failure”: Check your SSO credentials or API keys within the MCP server configuration.
  6. “Schedule Failed”: Ensure the service account has SELECT permissions on the underlying schema.
  7. “Missing Cost Metrics”: Verify that token usage and cost-per-token are passed as numeric fields, not strings.
  8. “SSO Login Loop”: Often caused by email mismatches between your IDP (Okta/Google) and Atlas user list.
  9. “Cohort Sync Error”: Check that your destination (e.g., Braze/Segment) has a valid API key in Settings.
  10. “Slow Report Loading”: Large unstructured JSON columns can lag; try extracting key AI properties into a flattened view.