WebMCP Challenge

Judge quick start

Browser setup

01Use ChatGPT’s desktop in-app browser, which supports WebMCP, or Chrome with chrome://flags/#enable-webmcp-testing enabled.
02Open the seeded demo. No account, credentials, or hardware are required.
03Open Available site tools and confirm exactly 12 BenchPilot tools.
04Run the prompts below and verify each change in the visible workspace.

All measurements are seeded or explicitly supplied by the judge. Resistance writes are rejected unless the power state is disconnected.

Prompt 1
“Inspect the active case and identify the safest next checks.”
Prompt 2
“Record the readings I just gave you: 32.1 and 31.8 ohms for the two disconnected coils.”
Prompt 3
“Update the hypotheses using only recorded evidence, then draft a cautious plan without claiming the motor is defective.”

Use Reset demo before a fresh run.

12 total

get_active_caseread

get_device_canvasread

list_measurementsread

list_hypothesesread

get_case_historyread

focus_canvas_entitieswrite

create_test_stepwrite

record_measurementwrite

update_hypotheseswrite

create_wiring_variantwrite

draft_diagnostic_planwrite

undo_agent_changewrite

Loading BenchPilot…