Reading usage

Track how much you use each LLM provider — calls, tokens, and estimated cloud cost. Usage is recorded per device in the desktop app and can sync to your org for admin reporting.

Open the usage dashboard

  1. In the desktop app sidebar, open Usage (or use the command palette: Ctrl+K → "Usage").
  2. The dashboard shows totals for the selected time range.

You can also see session-level token hints in the model toolbar during an agent conversation.

Time ranges

TabWhat it shows
This monthCurrent calendar month on this device
Last 30 daysRolling ~30-day merge of recent months
All timeUp to 24 months of local history

Click Refresh after a long agent run if numbers look stale.

Metrics explained

  • Total cost (USD) — estimated spend for cloud providers; Ollama contributes tokens at $0
  • API calls — number of LLM requests (agent turns, summaries, classifiers)
  • Total tokens — input + output tokens across providers

The per-provider table breaks down calls, tokens, and cost. Color cues highlight higher spend rows.

Export

Click Export CSV to write a usage report to disk. The toast shows the saved path — useful for expense reports or comparing Ollama vs cloud mix.

Settings budget bar

In Settings → AI Providers, cloud providers show monthly spend against an optional budget and warning threshold (warnAtPercent, default 80%). The usage dashboard is the detailed view; the budget bar is a quick guardrail.

What is not counted

  • Local MATLAB MCP tool execution (no LLM tokens)
  • Failed calls may still record partial usage depending on provider
  • Other users' devices (each desktop has its own local DB until sync)

Team admins see aggregated usage in the admin console; members see their own device dashboard.

Sync and privacy

Usage events sync when signed in and online. They do not include full chat transcripts unless your org explicitly enables content logging (uncommon).

Tips for solo engineers

  1. Start with Ollama — confirm cost stays near zero.
  2. Before a heavy cloud session, set a monthly budget in AI Providers.
  3. Export CSV at month end if you need to justify API spend to management.

Related

Reading usage | MBD Mind