Your next shopper is an agent. We build for it.

MerxonLabs builds tools, products and prototypes for agentic commerce in retail, for the moment customers stop browsing and send an AI agent to shop instead. Everything we test, we publish.

What we build →
simulation · illustrative scenario
briefweekly shop, 15 items · cap £60
changeoat milk goes out of stock
checksubstitute follows the customer's rule ✓
changedelivery fee rises to £5.50
checkbasket re-totalled, still under cap ✓
changepromo code expires at checkout
resultpurchase correct, 7 of 7 checks passed
What we build

Infrastructure for commerce run by agents.

Stores, catalogs and checkouts were designed for people looking at screens. We build the tools that test, measure and run commerce when the customer is an agent.

● Now building

A simulated retail environment for shopping agents

A controlled store where conditions change mid-purchase, so we can test whether shopping agents still complete the purchase correctly. Every run is scored against what the customer actually asked for.

  • Price changes
  • Stock-outs
  • Substitutions
  • Fees and delivery rules
  • Expiring promotions
  • Policy changes

Exploring next · ideas, not commitments

Agent economy analytics

Retail analytics for an economy of shopping agents: which agents visit, what they compare, what they buy, and why they leave.

Benchmarks and certification

An independent bureau that benchmarks, rates and certifies shopping agents on accuracy, budget discipline and honesty.

Commerce journey tester

Send agents through a store's full journey, from search to checkout, and see exactly where they get stuck or go wrong.

B2B procurement agent

An agent that buys supplies for a small business within explicit operational constraints: budgets, approved suppliers, stock levels and delivery windows.

Lab

Case studies and prototypes, in the open.

Each experiment is written up with its method, its counts and its limits.

  • 05.10.2026Prototype

    When the shop talks back to AI agents

    Parley, a store's representative agent, tested live on Ocado with Claude and GPT buyer agents.

    Read (PDF)
  • 29.09.2026Case study

    When AI agents do the weekly shop

    Claude and ChatGPT agents shop the same list at Tesco, Sainsbury's, Ocado and Asda.

    Read (PDF)
  • NextIn progress

    More experiments on the way

    New case studies and prototypes are published here as they're ready.

Method

Every product starts as a question.

01

Ask

A precise question about how agents shop, and what retailers gain or lose.

02

Test

Real agents on live stores, scored line by line.

03

Build

The smallest prototype that answers what the test exposed.

04

Publish

Write it up openly. The ideas that hold up become products.

About

Retail spent twenty years optimizing pages for people. The next customer reads differently.

A lab for agentic commerceWe focus on retail: the stores and brands whose products are now found, compared and bought by AI agents acting for shoppers.

Built in publicCase studies and prototypes go on this site as soon as they're ready, including the ones that didn't work.

Working with retailersWe're looking for retail and brand teams who want to explore what agent-ready commerce looks like for them.

MerxLatin for goods, merchandise. What every shopping agent is ultimately sent to find.

Contact

Building for agents? Let's talk.

Or write to us at contact@merxonlabs.com