Squad Can’t Delegate or Find a Model

When Squad cannot start a worker, check the role, agent, model, and session context in that order. Do not assume every delegation error means the selected model is invalid.

Check Settings

Select the intended computer.
Open Settings → AI tools and confirm the required agent is installed and connected.
In Squad, check Implementation, Review, or the two debate peers.
Select Refresh models.
Choose an available model or Agent default.
Select Save changes.

If a role is Ask when orchestrating, choose an agent for the current task or save a machine preference. An unavailable settings read is different from an intentionally unconfigured role.

Check the execution context

Run Squad from the agent user’s active Herdr session in the intended repository. A plain admin shell does not provide the same session context.

Command Code is not available as a native Squad agent. Model names and permission flags also differ across providers.

If a worker may already exist

Run squad status and inspect the original pane. A timeout or uncertain launch result can leave work running. Do not start a second worker until you know the first one’s state.

If authentication works but the model is denied, check provider entitlement, quota, and organization policy.

Report the useful evidence

Include the role, agent/version, selected model, catalog message, and redacted error. Avoid sharing full environment variables or credentials.

See Squad role configuration, command reference, and interrupted work.