š Open-Sourced: RiserFlow ā The AI Commerce Framework

Over the past few months at RiserLabs, weāve been building AI assistants for e-commerce. Along the way, we realized something crucial: the hardest part isn't the chat interface or the widget. The hardest part is teaching an AI agent to reliably interact with a real storeāfinding products, remembering customer preferences, managing the cart, and placing actual orders.
We extracted this core engine into a standalone open-source framework called RiserFlow.
What is RiserFlow?
Itās an MCP (Model Context Protocol) server that empowers any AI agent (ChatGPT, Claude, or your custom bot) to directly interact with your online store:
š Semantic Search: Syncs your catalog to a local DB, generates embeddings, and gives the agent a search layer far superior to legacy e-commerce APIs.
š§ Persistent Memory: The agent recalls customer details (size, color, favorite brands) across multiple sessions.
š Cart Management: Full control over the user's shopping cart.
š¦ Real Order Placement: Orders go straight into your actual CMS/admin panel, not a fake demo flow.
Not on Amazon. Not on a marketplace. Directly into your own store, your Bitrix, or any other engine.
Tech Stack & Architecture: Next.js 16+ ⢠Prisma ⢠PostgreSQL ⢠BullMQ ⢠Bitrix Adapter. Built with production in mind: includes Multi-tenancy (one server, multiple stores) and Idempotency (safe retries, no duplicate orders).
Why are we building this? We believe that in a couple of years, most customer-to-business interactions will happen via AI agents. Instead of clicking through filters and catalogs, users will simply say: "Find me the same t-shirt I ordered last time, but in black, and get it here by Friday."
The standards for this are just forming, and MCP is gaining serious traction right now. However, there is almost no ready-made MCP infrastructure for e-commerce. RiserFlow is our answer. An open framework that lets you connect an AI agent to a real store today. Not in a year. Now.
ā GitHub: github.com/riserlabs/riserflow Released under Apache 2.0 ā take it, fork it, contribute. Let the magic happen!