Investigate request latency
Investigate slow MCP requests in Durin Performance, compare gateway and provider timing, and check errors before retrying uncertain actions.
Inspect the timing
Open Performance in the Admin workspace.
- Select the time range containing the slow request.
- Identify its connection or tool and inspect the available timing breakdown.
- Compare Durin and upstream timing across similar requests.
- Match slow requests with errors or uncertain outcomes in activity.
Compare usage and cost
Use Performance for request volume, latency and data transfer, and Cost insights for modeled cost and reported usage. Select the relevant period and filters, then inspect chart details before comparing totals. Modeled cost estimates and provider-reported usage are distinct from Durin connection charges.
Troubleshooting
Most of the delay is upstream
Check provider state and errors for that connection. Verify with a small reviewed read after the provider recovers.
A slow write ended with an uncertain outcome
Check the upstream system directly before retrying. A timeout does not prove the write failed.
Frequently asked questions
Does one slow request prove the gateway is slow?
No. Compare similar tools and periods, then use the available breakdown to locate the delay.