Google

We recorded latency ranges from 233ms to 1249ms across different operations. Label listing and email reading completed in 233-277ms, while search operations and batch modifications took 1039-1249ms. Draft creation and label operations fell in the middle range at 450-709ms. Attachment downloads required 1249ms in our test environment.

Our tests used gmail.readonly, gmail.modify, and gmail.labels scopes. The readonly scope enabled email searching, reading, and attachment downloads. Label management operations required the labels scope, while email modifications and draft creation used the modify scope. We did not test with minimal scope combinations to determine the exact requirements for each tool.

Batch email modification through batch_modify_emails completed in 1238ms during our testing, while single email modification via modify_email took 776ms. We tested these operations independently but did not directly compare processing the same number of emails through both methods to measure efficiency gains.

Eight tools were skipped due to write-dangerous classifications in our test environment. These likely include operations like sending emails, deleting messages, or other irreversible actions. The specific skipped tools were not executed due to policy, dependency, or test-environment limitations rather than server capability issues.

Attachment download functionality executed successfully through the download_attachment tool, completing in 1249ms. This was among the slower operations we tested, but completed without errors. We did not test with various attachment sizes or formats to determine performance characteristics across different file types.

Label operations showed varied performance in our testing. Creating new labels via create_label took 450ms, while get_or_create_label required 709ms due to the additional lookup logic. Label listing through list_email_labels was faster at 240ms, and updating existing labels completed in 665ms.

Email search functionality executed successfully via search_emails in 1039ms, and filter listing worked through list_filters in 233ms. Filter listing was among the fastest operations tested. We did not test complex search queries or filter creation capabilities, focusing on basic search and filter retrieval operations.