Cartwheel Grocery
grocery · 5 audit runs · task: find a product and add it to the cart · sample data (illustrative fixture, not a real measurement)
Readiness score
19/100
70% task success · 15% WebMCP tool surface · 15% efficiency (steps to complete). Full rubric in the open methodology.
The WebMCP delta
Agent success via raw UI
20% · avg 15 steps
Agent success via WebMCP tools
No WebMCP tools exposed — this is the opportunity.
Why runs fail here
timeout2 of 4
auth wall1 of 4
missing info1 of 4
Run log
FAIL ui 18 steps 90.2s timeout
FAIL ui 20 steps 103.8s auth_wall
FAIL ui 12 steps 91.1s missing_info
PASS ui 15 steps 48.0s
FAIL ui 19 steps 126.5s timeout