Reports
The Performance Report is your workforce's monthly review, written from live data. It answers the questions a manager actually asks: how much did my agents get done, how well, how fast, at what cost, and how safely? Every figure comes from your real workspace activity — nothing is mocked.
You reach it from the sidebar under Monitor → Reports.
The report is readable by everyone on the team, including Viewers. It's designed to be shared — copy a summary into Slack, or export a PDF for a stakeholder who doesn't have a login.
Choosing a period#
Top-right, pick the reporting window: 7 days, 30 days, or 90 days. Every section — the metrics, the charts, the agent breakdown — recalculates for the range you choose. The date range is shown under the title.
The executive summary#
The report writes its own opening paragraph. Based on the live numbers it produces a headline (for example, "Your AI workforce delivered 1,204 actions, the equivalent of ≈ 3.2 human workdays") and a short bulleted briefing calling out the top-performing agent, the share of work that succeeded without intervention, and — if any — how many decisions are waiting on you.
The metrics strip#
A single row of headline numbers sits below the summary:
| Metric | What it means |
|---|---|
| Actions | Total actions completed in the period, with a daily average |
| Success rate | Share of finished work that succeeded (green at 90%+) |
| Time saved | Estimated manual work avoided, versus doing it by hand |
| Approved | Approvals your team cleared |
| Rejected | Approvals your team sent back |
| Pending | Approvals still awaiting review |
| Avg wait | Average time a decision sits before someone acts on it |
The sections#
The body of the report is numbered like a document:
- Output — a trend of daily actions completed, active days, the peak day, and total time saved.
- Agents — one "payroll receipt" card per agent (swipe/scroll horizontally): actions completed, share of output, 7-day success rate, and hours of human work saved. Each links through to that agent's detail view.
- Decision quality — the split of outcomes into Worked, Needs review, and Failed, plus how many approvals you approved versus rejected and how many things your agents learned this week.
- Responsiveness — how quickly agents respond after an event fires, shown as a median (P50) and slowest-5% (P95) line. Gaps in the line are days with no work claimed.
- AI usage cost — total spend, tokens, and LLM calls for the period, broken down by model.
- Safety — guardrail enforcement: actions Blocked, Approval-gated, Auto-paused, and Edited approvals, plus your correction rate.
- Insights — automatically surfaced patterns across your agents' work, each with a recommendation and a link to act on it.
On cost attribution: the report notes that per-agent cost isn't broken out yet — all agents currently share one API key, so cost is reported workspace-wide. If you bring your own provider keys, spend routes to your own accounts.
Exporting and sharing#
Click Export (top-right) for four ways to take the report with you:
| Option | Format | Use it for |
|---|---|---|
| Daily breakdown (CSV) | CSV | A day-by-day table of actions, time saved, and approvals — for spreadsheets |
| Agent summary (CSV) | CSV | Per-agent totals, share of output, and success rate |
| Copy team update (Slack) | Clipboard | A paste-ready digest for a standup or Slack channel |
| Download PDF report | A formatted document of the whole report for stakeholders |
The Copy team update option puts a tidy, emoji-tagged summary on your clipboard — actions completed, workdays saved, success rate, top performer, and approval counts — ready to paste into a message.
CSV exports are safe to open in Excel or Google Sheets — Syntrum neutralizes any cell that could be misread as a spreadsheet formula.
What's next#
- Audit log — go from an aggregate number to the individual events behind it.
- Approvals — clear the pending decisions the report flags.
- Models & BYOK — bring your own keys so usage cost bills to your accounts.