BLAZE MCP Connects AI Assistants to Cannabis Retail Operations

BLAZE on September 29, 2026 launched BLAZE MCP, a Model Context Protocol server that connects Claude, ChatGPT, and other MCP-compatible AI assistants to more than 250 operations across the company’s point-of-sale, e-commerce, and marketing products for cannabis retailers. The launch announcement, datelined Anaheim, California, describes the product as the most comprehensive MCP server built for cannabis.

Model Context Protocol is an open standard that lets AI assistants work with business data rather than only converse about it, according to BLAZE’s BLAZE MCP product page. BLAZE said most AI tools in cannabis retail can see one system, while BLAZE MCP reaches the whole store because the company builds point of sale, e-commerce, and marketing on one connected platform. Under that design, the company said, a single conversation can reach pricing and inventory at the register, the online menu, in-store screens, and the campaign that brings customers back.

BLAZE said more than 140 of the server’s operations are actions that make real changes rather than lookups. Operators can mass-edit up to 50 products at once, set VIP and loyalty-tier pricing, receive purchase orders, reconcile batches, redraw delivery zones, and diagnose why a product is not showing online or why an order is stuck, according to the company.

The company’s example workflow: a store manager can ask which products are moving slowly, discount one while protecting a 30% margin, build the promotion, feature it across every online storefront and in-store displays, and launch a campaign announcing it, all without leaving the conversation. BLAZE said the combined workflow turns “a multi-tool afternoon into minutes.”

“Every cannabis software company is talking about AI. The question operators should ask is what it can actually do,” Chris Violas, CEO of BLAZE, said in the announcement. “BLAZE MCP does the work across the entire store because every piece of it runs on one platform. That depth and breadth is what separates a chatbot from a real member of the team.”

How the Connector Works

The product page lays out a four-step flow. In the Connect step, the user links a preferred AI assistant to BLAZE, scoped to that user’s existing role and permissions. In Prompt, the user states the task in plain language. In Review, the system spells out exactly what is about to change before anything happens. In Execute, the user approves and the change goes live, logged in full.

Listed use cases include asking an assistant to flag the slowest-moving SKUs from point-of-sale inventory and draft a discount that keeps at least a 30% margin; turning that discount into a smart collection merchandised with a banner across every store; drafting a one-time outreach campaign announcing the promotion; and finding frequent shoppers who are not yet enrolled in the loyalty program and drafting invitations to sign them up.

The company also lists the assistant’s scope by product line. For point of sale, it can query live inventory, adjust prices, check compliance, and track core store operations. For e-commerce, it can sync digital storefronts, merchandise online stock, and keep online menus aligned with the shop floor. For marketing, it can analyze customer sales trends, build segmented audiences, and trigger targeted promotional campaigns. For loyalty, it can access reward profiles, apply points, create targeted discounts, and automate personalized customer retention offers.

Permissions, Logging, and Availability

BLAZE said every change the assistant proposes is previewed and requires human approval before it takes effect, and each step of a multi-step workflow is confirmed individually. The assistant can only access what the user’s existing BLAZE role allows, every action is recorded in a full audit trail, and customer data is never used to train AI models. The product page adds that assistants connect with zero shared credentials, using the user’s authenticated BLAZE profile, and that every request and system action is logged instantly.

BLAZE said it designed and built the AI layer in-house on its SOC 2 Type 2 certified platform. The product page’s frequently-asked-questions section states that nothing is applied automatically: if a draft discount, banner, or campaign is not right, the user edits or rejects it before it goes live.

BLAZE MCP is available now as an add-on for BLAZE customers, and the product page notes that availability depends on the customer’s subscription level. The company is also inviting operators to join an early access program. Setup is permissions-based inside an existing BLAZE account, with no integration to build or server to stand up; users search for the BLAZE MCP connector inside their preferred MCP-compatible assistant and connect it from there. For accounts that span multiple store locations, the assistant can work across all of them in the same conversation, scoped to whatever locations the user’s role permits.

BLAZE describes itself as an AI-powered cannabis retail platform trusted by thousands of businesses across 20 states.