With Port Execution Agent that use polling streamer (https://docs.port.io/actions-and-automations/setup-backend/webhook/port-execution-agent/usage#polling-streamer), messages are acknowledged before processing and cannot be claimed by another agent once fetched.

This allows multiple agents to run in parallel without conflicts, but requires using the same filtering logic across all agents to avoid unintended filtering.

You cannot control which agent processes which message at the moment.

Suggested solution:

Set and optional "cluster identifier" that is evaluated when executing the run. The poller will send its "cluster identifier", only matching runs will be fetched.