Usage
Track AI usage across your workspace and your personal account — totals, breakdowns, time-series charts, and CSV export.
The Usage pages give you visibility into how Opisense's AI is being used. Every assistant message, AgentForce conversation, knowledge-base query, and designer turn flows into the same metrics pipeline, so you get one consistent view across the whole product.
There are two pages:
/settings/workspace/usage— Admin-only. Workspace-wide totals plus per-user, per-agent, and per-model breakdowns./settings/personal/usage— Every member sees their own usage.
What's tracked
For every AI interaction (assistant turn, AgentForce reply, KB chat answer, designer turn, heartbeat tick), Opisense records:
| Field | What it captures |
|---|---|
| Agent / surface | Which agent or surface ran — assistant, customer-support, sales, kb-chat, designer, etc. |
| User | Who triggered it |
| Model | The AI model that produced the response |
| Input tokens | Prompt tokens sent to the model |
| Output tokens | Completion tokens produced |
| Errors | Whether the run failed |
| Cost | USD cost (ships in a follow-up; currently $0.00) |
These events roll up into daily and monthly totals, which power the dashboard.
The dashboard
Both pages share the same shape:
Totals cards
At the top, a row of summary cards shows:
- Agent runs — Total runs in the selected period
- Tokens (input + output) — Aggregate token consumption
- Errors — Count of failed runs
- Active members (workspace only) — Unique users who ran something
The selected period defaults to the current month. Pick a different range from the date picker to look further back.
Time-series chart
A stacked area chart shows agent-run volume over time, broken down per agent or per model (toggle the breakdown axis). Hover any point to see the exact counts.
Breakdown table
Below the chart, a table lists per-agent or per-user totals for the selected period, with columns for runs, tokens (input/output), errors, and (eventually) cost.
The table sorts by runs by default; click any column header to re-sort.
CSV export
Both the workspace and personal usage pages have a Download CSV button in the page header. The CSV contains every row from the breakdown table for the selected period. Useful for:
- Importing into a spreadsheet for finance reviews
- Building your own dashboards in BI tools
- Archiving usage snapshots month-over-month
The CSV format follows the same columns as the on-screen table.
Permissions
- Workspace usage is admin-only. Members trying to access
/settings/workspace/usagesee an "insufficient access" message. - Personal usage is available to every member, but only shows their own data.
How metrics get there
Usage metrics flow from Mastra (the AI agent layer) into Convex (the backend) via a /usage/ingest endpoint. The flow is asynchronous — events are buffered for ~30 seconds before being flushed in batches.
This means:
- Brand-new activity may take up to 30s to appear on the dashboard
- If Mastra is down, events are buffered until it's back; you won't lose data unless the buffer overflows
- Rollups (daily, monthly) are computed in Convex from the raw event stream
You don't need to manage any of this; it just runs.
Cost (cost in USD)
In v1.0, the Cost column shows $0.00 across the board. Cost-pricing integration is a follow-up — Mastra's pricing registry isn't wired up yet. Once it lands, cost values will populate retroactively for any events still in the rollup window (typically the current month).
Until then, tokens are the most reliable signal for usage size. Heavier usage = more tokens.
Plan limits
Plan limits (e.g., monthly token caps, per-user caps, per-agent caps) are enforced separately from this dashboard. If you're hitting limits, you'll see them surfaced in the relevant feature (e.g., heartbeat ticks marked as daily_token_cap, KB chat showing a quota warning).
For plan details, see Billing.