A middleware that sits between your customers and your Magento 2 GraphQL API. It translates natural language queries into precise product filters — no Composer packages, no deployments, no version conflicts.
5 minutes to connect. Works with any Magento 2.4+ store.
ID: 43
ID: 28
0-100
ID: 52
ID: 167
Three steps. Five minutes to first query.
Create an integration token in Magento Admin. Give it to the middleware. No Composer, no deployment, no SSH.
Admin > System > Integrations > Add New. Takes under 5 minutes.
The middleware queries your store's GraphQL API once and caches the full schema: every filterable attribute, every option value, every price range.
Uses products(search:"", pageSize:1) aggregations + customAttributeMetadata.
Natural language queries are matched against your cached attribute options and converted to precise GraphQL filter objects with resolved option value IDs.
Choose LLM mode (Claude API) for complex queries or regex mode for zero cost.
The middleware layer between natural language and Magento GraphQL.
Handles ambiguous queries, synonyms, typos, and complex multi-attribute filters. Best for fashion, electronics, and stores with large catalogs.
Pattern matching against cached attribute options. No external API calls. Best for stores with predictable, structured queries and fewer attributes.
Any Magento 2.4+ store that wants natural language product search without touching their codebase.
Complex attribute combinations: brand, color, size, material, style, season. Customers search naturally — the middleware resolves it all.
Spec-heavy product catalogs with precise technical filters. Turn natural language specs into exact attribute matches.
Thousands of SKUs, hundreds of filterable attributes. The middleware auto-discovers everything and scales with your catalog.
How the middleware discovers and translates your store schema.
A single query returns every filterable attribute with available values and counts. No catalog export. No manual mapping.
products(search: "", pageSize: 1) {
aggregations {
attribute_code
label
options {
label
value
count
}
}
}Cached on startup. Refreshed on schedule or manually.
Discovers the full store schema: attribute types, input types, and entity types for every custom attribute.
customAttributeMetadata(
attributes: [{
attribute_code: "color"
entity_type: "catalog_product"
}]
) {
items {
attribute_code
attribute_type
input_type
}
}Resolves attribute types for proper filter construction.
The parser matches natural language tokens against cached attribute options, resolves to internal option value IDs, and constructs a valid GraphQL filter object.
products(
filter: {
brand: { eq: "43" }
price: { from: "0", to: "100" }
color: { eq: "52" }
size: { eq: "167" }
}
sort: { relevance: DESC }
pageSize: 20
) {
items {
name
sku
price {
regularPrice {
amount { value currency }
}
}
}
}Filter values are option IDs, not labels. The middleware handles the resolution.
Minimal requirements. Maximum compatibility.
Open source core. Managed service for teams that want zero ops.
Full source code. Run it yourself.
Per store. We host and operate it.
Multi-store. SLA. Dedicated support.
Zero install opens the market from 'stores with a dev team' to 'any Magento 2.4+ store'.
"We connected the middleware to our store in under 10 minutes. No Composer, no deployment, no version conflicts. It just discovered our catalog and started translating queries."
Common questions about the middleware.
No. The middleware only needs a Magento integration token with catalog read access. It connects via the standard GraphQL API. No Composer packages, no code changes, no deployment.
LLM mode (Claude API) handles complex, ambiguous queries with synonyms and typos. Regex mode matches patterns against cached attributes at zero cost. You can use both: regex for simple queries, LLM as fallback.
One GraphQL query (products with aggregations) returns all filterable attributes, their options, and counts. customAttributeMetadata adds type information. Cached and refreshed on a schedule.
Magento 2.4+ and Adobe Commerce 2.4+ are fully supported. Magento 2.3.x has limited support since some GraphQL aggregation features were added in 2.4.
A valid Magento GraphQL products query with resolved filter objects. Filter values are internal option IDs, not labels. The response includes the matched products with all requested fields.
Yes. The parser interface is pluggable. Claude API is the default and recommended option, but you can integrate OpenAI, a local model, or any API-compatible LLM.
Share your store details. We'll reply with a technical assessment and connection instructions.