Magendoo / AI Lab

Working software. Try it.

The Lab is where our agentic commerce tooling runs in the open: agents searching a Magento catalog in natural language, gated write operations with audit trails, and operational intelligence for running stores. No slideware — exhibits below.

3 exhibits · built on MageMCP & open-source tooling · github.com/magendooro

Ask a store for products. In plain language.

Exhibit 01

Agentic Search translates natural language into Magento 2 GraphQL against a live store — it discovers the store's own attributes, option values and category tree, then resolves the query against them. Zero install on the store side.

Agentic Search — demo Magento 2.4 store Recorded sessions · live sandbox coming

1 — Intent resolved against the store's discovered schema

2 — Generated GraphQL


            


            

                
3 — Live results from the demo catalog

                

            

        

    

    

        Replays of real middleware sessions against a demo catalog. Store data is illustrative.
         → How Agentic Search works
    







    

        
An agent operating a store — with the safety on.

        
Exhibit 02

    

    

        MageMCP gives AI agents 60 typed tools over Magento 2 / Adobe Commerce. Reads are free;
        writes stop at a confirmation gate and every invocation lands in an audit log. The replay
        below shows a representative session — real tool names, real gating, real audit format.
    

    

        

            MageMCP session — inventory restock
            
        

        

    

    

        Tool names and confirmation semantics (confirm=true on second call) match the shipped server.
        60 typed tools · 895 automated tests · MIT.
         → Source
         · All MCP servers
    







    

        
Agents that watch the store, not just the shopper.

        
Exhibit 03

    

    

        

            
MageIQ — analytics over TelegramSample

            

                
Ask your store a question.

                
Plain-language analytics against live Magento data.

                
YouHow did we do yesterday?

                
MageIQYesterday: 214 orders, €18,940 revenue (+12% vs. same weekday last month). Top category: Running Shoes. 3 products went out of stock — RS-042 had 41 views after stockout.

                
YouWhich of those should we restock first?

                
MageIQRS-042 — highest demand while unavailable and a purchase order is already open. RS-017 and BT-203 sold under 2 units/week.

                
→ About MageIQ

            

        

        

            
StoreSignals — operational alertsSample

            

                
Know what broke before customers tell you.

                
Signal detection across catalog, search, checkout and fulfillment.

                
HIGHCheckout error rate 2.1% → 6.8% in the last hour. Pattern: payment gateway timeouts, 12 occurrences.

                
MED184 products in "Winter Boots" have no image on the configured storefront — category filter renders them last.

                
MEDSearch "gore tex" returns 0 results; 37 searches this week. Synonym for "gore-tex" missing.

                
INFONightly ERP stock sync completed with 0 mismatches (14,208 SKUs).

                
Sample alert feed — illustrative data. → About StoreSignals

            

        

    







    
Under the hood

    
Every exhibit is inspectable.

    

        
Agentic Search
Middleware translating natural language to Magento 2 GraphQL — discovers each store's attributes, option values and price ranges at connect time. Zero install on the store. Details →

        
MageMCP
Open-source MCP server for Magento 2 / Adobe Commerce: 60 typed tools (18 storefront via GraphQL, 42 admin via REST), 895 automated tests, PII redaction, write confirmations, audit logging. Source →

        
MageIQ & StoreSignals
Operational agents on the same engineering discipline: typed tools, permissions, and audit trails — not a chatbot bolted onto an API key. MageIQ → · StoreSignals →

    







    
This is what the assessment designs for your store.

    

        The exhibits run on our demo store. The Agentic Engineering Assessment maps what the same
        discipline — typed tools, gated writes, measured outcomes — looks like inside your
        codebase and your controls.
    

    

        Start an Agentic Engineering Assessment
        Prefer email? [email protected]