Skip to content

Token usage and Auto mode

The Token usage panel shows your real consumption, measured locally, and refreshes itself as each chat turn closes. It has three scopes:

  • Current project — how much the open project has consumed: tokens today, this week and this month, plus the last 7 days as a series.
  • Global — usage across the whole app, with a “top” list of the projects that spend the most.
  • Providers — quotas and limits reported by each provider (Claude, OpenAI, Z.AI, MiniMax…). It only shows figures the provider hands over: if a plan has no public quota API, the panel says so plainly instead of inventing an estimate.
Token usage: today, week, month and last 7 days, per project and global.

Auto mode: profile rotation without stopping the task

Section titled “Auto mode: profile rotation without stopping the task”

Auto mode keeps work moving when an engine fails or runs dry: if an agent’s preferred provider is unavailable — context full, tokens exhausted, quota, authentication down or saturation — the app rotates to the next profile on your list without halting the task.

  • Turn it on with the Auto toggle and reorder the profiles by dragging them: that’s the fallback order the orchestrator and its subagents will use.
  • Each agent tries its own configured provider first; fallback only kicks in when that one fails.
  • The “Default order” button restores the original list.
The per-provider breakdown tells you where every spent token came from.
The system monitor watches CPU, RAM, GPU, disks and network while local models work.