Loading workspace…
Contexts
What agents report using your tools for.
Weather MCP · Synthetic context cohort
Usable context reports8,960Of 12,840 recorded tool calls
Context coverage69.8%3,880 calls have no usable report
Stable themes77,730 calls assigned to a theme
Unassigned reports1,230Available to explore · no review required
| Theme | Calls ↓ | Share | Δ share | Tool called most | Open |
|---|---|---|---|---|---|
| “Will it rain during our hike?” | 2,300 | 25.7% | — | ||
| “Which destination has better weather?” | 1,900 | 21.2% | — | ||
| “Find a clear spot for stargazing within a two-hour drive.” | 1,480 | 16.5% | +6.4 pp | ||
| “Do I need a coat this evening?” | 920 | 10.3% | — | ||
| “Are storms expected before departure?” | 620 | 6.9% | — | ||
| “Choose a dry window for tomorrow’s site visit.” | 330 | 3.7% | — | ||
| “Check conditions at our usual location.” | 180 | 2.0% | — |
Share is of all 8,960 usable reports in this tool scope, including 1,230 unassigned reports. Themes are inferred. Δ share uses the same theme definition; unavailable comparisons are withheld.
Themes learn automatically
New reports are assigned to stable theme IDs first. Distinct, supported patterns become new themes; emerging and weak patterns stay inspectable. Review is optional. This describes fixture behavior; no model ran.
Separate synthetic Weather MCP cohort. Agent reports are unverified. Counts are calls, not unique users. Themes are inferred separately from captured text. Task outcomes and missing conversation identifiers remain unknown.