Catalogian is designed for reliable, continuous monitoring of your data sources. This page covers our uptime targets, how checks are scheduled, and what happens when things go wrong.
Catalogian targets 99.9% uptime for the REST API, MCP, and OpenAI Responses endpoints. This covers the API being reachable and returning valid responses — not the availability of your upstream sources.
Monitoring checks run on the interval you configure (e.g. every hour). Checks are scheduled relative to the last completed check, not wall-clock time. If a check takes 30 seconds to process a large source, the next check starts your configured interval after that check completes.
You can optionally set a check anchor time (e.g. 06:00 UTC) to align checks to a specific time of day. This is useful when your upstream publishes new data at a known hour.
If Catalogian can't fetch or parse your source, the check is recorded as a failure. Common causes:
| Failure type | What happens | Next check |
|---|---|---|
| HTTP error (4xx/5xx) | Check marked as failed, no delta event created | Retries on next scheduled interval |
| Connection timeout | Check marked as failed after 60s timeout | Retries on next scheduled interval |
| DNS resolution failure | Check marked as failed | Retries on next scheduled interval |
| Invalid file format | Check marked as failed with parse error | Retries on next scheduled interval |
| Response too large (500 MB free; up to 5 GB staged on paid records) | Check rejected, error recorded | Retries on next interval (will fail again until the response shrinks) |
Failed checks do not count against your delta retention. They are visible in the source's activity log in the dashboard.
Catalogian continues to retry on your configured schedule even after multiple failures. Sources are not automatically paused due to failures. If a source has been failing for an extended period, you'll see the error in the dashboard and can investigate.
Set up webhooks for monitoring. While Catalogian doesn't send webhooks for no-change events, you can use the get_health MCP tool or the REST API to check source health programmatically.
| Record tier | Delta history window | Rows |
|---|---|---|
| Free | 7 days | Never pruned to the window |
| Watch | 30 days | Never pruned to the window |
| Max | 180 days | Never pruned to the window |
The window applies to delta history: after it closes, delta events past the edge are permanently deleted. The record's current rows are never pruned to the window; they stay for the life of the record.
Maintenance windows are scheduled during low-traffic periods and announced in advance. During maintenance, the API may return 503 responses briefly. Monitoring checks that fall during a maintenance window are rescheduled automatically.
If you're experiencing issues with Catalogian:
Compare plans and features. Plans & Limits →