AI Commerce Evals
Automated shopping benchmarks across stores and AI models. Multi-turn sequences, cross-model comparison, and PDF reports. Built on the same engine that powers the agent benchmark and leaderboard.
How It Works
Define
Pick stores, models, and write a multi-turn sequence. Generic prompts work across any catalog.
Run
Every store x model x sequence combination runs as an independent multi-turn session. 2 stores x 2 models = 4 sessions.
Report
Get a funnel comparison matrix, performance table, error classification, and a downloadable PDF report.
Example Sequence
Each turn is a separate API call. The agent sees full conversation history. Up to 8 internal tool-calling turns per message.
Sample Results
| Store | Model | Outcome | Tokens | Duration |
|---|---|---|---|---|
| oakywood.shop | Gemini 3.5 Flash | checkout reached | 154,294 | 34.7s |
| oakywood.shop | Gemini 3.1 Pro | checkout reached | 85,614 | 93.4s |
| ugmonk.com | Gemini 3.5 Flash | checkout reached | 96,058 | 29.9s |
| ugmonk.com | Gemini 3.1 Pro | checkout reached | 46,084 | 35.1s |
Use Cases
For Store Owners
Run evals against your own store to verify your UCP implementation handles multi-item carts, variants, and checkout flows correctly.
For Agent Builders
Compare models head-to-head on the same stores. Find which model reaches checkout fastest and cheapest.
For CI/CD
Trigger eval runs on every deploy. Assert on checkout rate, error count, and duration thresholds. Catch regressions before production.
Headless API
curl -X POST -H "Authorization: Bearer $TOKEN" \
-H "Content-Type: application/json" \
https://ucpplayground.com/api/v1/collections \
-d '{
"name": "Q1 Benchmark",
"config": {
"stores": ["oakywood.shop", "ugmonk.com"],
"models": ["gemini-3-flash", "gemini-3-1-pro"],
"sequences": [{
"name": "browse_and_buy",
"turns": [
{ "message": "Show me two products under $60" },
{ "message": "Add both to my cart" },
{ "message": "Proceed to checkout" }
]
}]
}
}'curl -X POST -H "Authorization: Bearer $TOKEN" \
https://ucpplayground.com/api/v1/collections/{id}/runcurl -H "Authorization: Bearer $TOKEN" \
https://ucpplayground.com/api/v1/collection-runs/{run_id}Learn more in the Evals & Collections help docs.