Launch field reference

Field	Purpose	Default guidance
Name	Human-readable identity for the agent and provider resource naming.	Keep the generated name or use a purpose-specific name.
Cloud Provider	Infrastructure provider for the VM.	Use the recommended enabled provider.
Runtime Profile	CPU everyday runtime or GPU voice and open-model runtime.	Use CPU unless you need GPU capabilities.
Region	Geographic provider region or automatic placement.	Use auto placement unless residency or nearby resources matter.
Zone	Failure/capacity zone inside a region.	Use any unless a dependency requires a zone.
Instance Size	Provider machine type with vCPU and RAM.	Start with the recommended tier.
Boot Disk (GB)	Persistent runtime and data storage.	Keep the enforced minimum; increase for code or media.
Exposure	Internet allowlist or Tailscale private access.	Use a narrow internet allowlist or Tailscale.
Allowed IP ranges	CIDRs permitted to reach a public owner UI.	Add the current IP as /32.
Microsoft Teams Funnel	Public Teams Activity endpoint for a Tailscale agent.	Enable only for the Teams bot transport.
LLM Credentials	Managed OpenRouter or BYOK.	Use managed for fast setup; BYOK for direct provider billing.
OpenRouter Attribution	Sends Pneum.ai app attribution on managed requests.	Leave enabled unless policy requires otherwise.

Dependencies

Provider changes reload locations and machine types. Runtime changes can reload GPU availability and raise the disk minimum. Region changes reload zones. Exposure changes reveal either IP or tailnet controls. Managed credentials reveal attribution settings.

Validating a selection

Wait for location, machine, and estimate requests to finish. Read any availability reason under the field. Confirm the estimate configuration matches the visible selection before launch.

For explanations and tradeoffs, see launch options.