Loading workspace…
Discoverability
Run prompt simulations to see when agents call your MCP and what happens next.
B04 · Latest batch
- MCP use rate
- 67% 12 of 18 runs called a tool
- Tool calls
- 29 Across 3 MCP tools
- Tool errors
- 0 Returned errors · 0% of calls
- Tracked prompts
- 3 1 prompt groups · 6 runs per prompt
MCP use over time
Latest batch only100%
50%
0%
67%
B04 · LatestEarlier batches are unavailable for this selection. MCP use counts runs with at least one tool call; task success is unknown.
Which configurations call your MCP?
Same prompts · 2 repeats per configuration · latest batch
By modelRuns with an MCP call
By harnessRuns with an MCP call
These harnesses use different models. Run the same model in both harnesses to compare their effect.
Where does discoverability vary?
Latest batch · runs with an MCP call / total runs
Less useMore use
3 models selected
| Prompt group | Model AHarness A | Model BHarness A | Model CHarness B | All configurations |
|---|---|---|---|---|
| Exports3 prompts | 66.7%4 / 6 runs | 83.3%5 / 6 runs | 50%3 / 6 runs | 66.7%12 / 18 runs |
Prompt group totals
MCP use by prompt group · a run counts once, even when several tools are called.
| Prompt group | MCP use rate | Used / runs | Prompts | Tool activity |
|---|---|---|---|---|
| Exports | 66.7% | 12 / 18 | 3 | Find records and create files |