Agent builders
Ship your shopping agent with a test suite, not a prayer. Run it against 50 shop variants in CI before every release.
Spin up realistic, disposable online stores via API. Test your shopping agents against carts, checkouts, and edge cases — before they touch real orders.
curl -X POST https://api.shopdojo.dev/v1/shops \
-H "Authorization: Bearer sk_test_..." \
-d '{
"archetype": "fashion-d2c",
"chaos": ["out_of_stock_at_checkout", "expired_coupon"],
"seed": 42
}'{
"id": "shop_8Xk2m",
"url": "https://shop-8xk2m.shopdojo.dev",
"state": "ready",
"reset_url": "https://api.shopdojo.dev/v1/shops/shop_8Xk2m/reset"
}Right now, you're testing your agent in production. Every failed run is a real cart, a real order, a real apology email.
One API call. Realistic catalog, cart, multi-step checkout. Deterministic by seed.
Inject the failures that kill agents in the wild: out-of-stock at checkout, address validation loops, expired coupons, consent walls.
Every click, cart change, and checkout step as a structured event stream. Same seed, same shop, same result — every run.
Ship your shopping agent with a test suite, not a prayer. Run it against 50 shop variants in CI before every release.
Building “buy for me” into your product? Validate the full purchase flow against edge cases your PM hasn't thought of yet.
Benchmarks need worlds. Deterministic, resettable shops make agent evals reproducible instead of anecdotal.
Deterministic worlds. Reproducible runs.
deterministic · seeded rng · resettable
Chaos on demand — the failures production would bill you for.
failure modes · injectable via api
Not a QA service. Not another eval framework. Not a browser sandbox. ShopDojo is the world your tests run in — bring your own agent, your own evals, your own browser.