Self-improving commerce · Cursor Commerce London

The storefront that
improves itselffor humans and AI agents.

Darwin watches how people and shopping agents move through your store, finds where they drop, changes the page, proves it with an A/B test and ships the winner as a pull request. Then it does it again.

Open mission controlVisit the demo storeSet up your store
behaviour→insight→page change→A/B test→pull request↺
Generation
0
Observing
Autopilot
Observer
Watching 1,204 humans and 318 AI agents shop PACE

Two audiences, one loop

Humans get a better page. AI shoppers get a better agent API: stock, delivery ETA, returns, landed price and negotiation, over MCP and A2A.

Safe, declarative changes

Every change is a validated PageSpec diff: A/B tested, reversible, and shipped as a normal pull request your team reviews.

Proof, not vibes

Bayesian A/B tests decide. Losers are rejected, winners become the next generation. Simulated traffic is always labelled.