NSL returns a ranked list of items for each request, and records which list a person saw so that later events can be credited to it.
How a recommendation is made
For predictable placement behaviour, send a context with each request: one placement in your product, such as a homepage feed or a “similar items” rail. The context decides which model retrieves candidates, how they are ranked, and which business rules apply. Operators change a context in the console; your code keeps sending the samecontext_id. See Key concepts for every term on this page.
Choose how to start
Quickstart (API)
Create credentials, send an item and an event, and fetch recommendations in about ten minutes.
No-code start
Paste your website’s address. NSL builds the catalogue from your pages.
Website embed
One script tag renders recommendations and records views and clicks for you.
Shopify, Magento and other platforms
Install the app or module for your shop platform.
What is in these docs
Integrate
Credentials, catalogue, events, recommendations, search, history imports and identity.
Console
Every page of the console, in the order the sidebar shows them.
Playbooks
Step-by-step tasks that combine several features.
SDKs
JavaScript, PHP, Next.js and Swift clients, and the MCP server.
API reference
The Core API for your application and the Platform API for automation.
Changelog
What changed, by product area.
For background on how recommendation systems work in general, read the recommendation systems reading path.