Limits and scaling
For: cluster operators and plugin authors.
Plan capacity from the tenant limits, the queues your plugins need and the measurements for your deployment. Agent counts alone do not describe the work a node or Temporal backend must handle.
A tenant's agents run on Temporal task queues, and every queue a node polls is registered in the tenant's Worker Deployment version when a new runtime version first polls it (Temporal). That registration is what limits how many queues a tenant may have, so Orchestrator Zero keeps the queues few: a plugin's agents and flows share one queue for each set of node labels they require (routing).
Check Compatibility for platform requirements and integration maturity, and Known issues for operational problems and workarounds.
Related tasks and reference
Previous section links
10,000 agents in one tenant
a fleet under load: Read the section.
a tenant of many small plugins: Read the section.
how many queues a plugin uses: Read the section.
more than the limits: Read the section.
near the limit: Read the section.
other temporal backends: Read the section.
run it yourself: Read the section.
the limits: Read the section.
what a pool means for capacity: Read the section.
what limits what: Read the section.
what we measured: Read the section.
when a plugin update adds an agent: Read the section.