multi-brand and multi-store: how a marketplace handles it all with one agent.
heysteff is an AI platform for customer support and sales across WhatsApp, Instagram, Messenger, Gmail and Shopify; Steff is the AI agent that runs it. A marketplace doesn't have a single brand tone or a single catalog: it has several, sometimes across several countries. heysteff is multi-tenant from the ground up, not a single agent forced to simulate multiple identities.
One workspace per brand or per country
Each brand — or each country, if you operate in more than one — lives in its own workspace: its own knowledge base, its own trained tone, its own conversation inbox. It isn't one generic agent answering on behalf of every brand alike; each workspace responds with the identity it belongs to, without mixing catalogs or policies across brands.
This separation is RLS-first, not a folder-organization convention: one workspace's data is neither visible nor accessible from another. For a marketplace that grants access to different teams — sometimes external, sometimes from the brand itself — that real separation is the difference between an auditable operation and one where anyone could, by mistake, see data from a brand that isn't theirs.
Roles, permissions and audit logs per team
With configurable roles and permissions, each person on the team sees and operates only what belongs to them: one brand's operator doesn't need access to another's inbox. Every action lands in an audit log, so in an operation with several teams working in parallel there's always a trace of who paused a bot, changed a setting or took over a conversation.
For the marketplace's central team, this replaces the manual process of asking each brand for a report of what changed: the audit log is already there, workspace by workspace, ready to review whenever needed without depending on each brand documenting it on its own.
Channels per brand, never crossing over
Each brand connects its own channels — its WhatsApp number, its Instagram account, its Messenger, its Gmail — inside its own workspace. A customer who writes to brand A never ends up in brand B's inbox, even though both run on the same platform.
Up to 3 Shopify stores on the same plan
The Max plan includes up to 3 connected Shopify stores, each with its own abandoned cart recovery and its own catalog — useful for a marketplace running several Shopify stores under the same umbrella, or a brand with separate stores per country or product line.
Routing and assignment per team
Within each workspace, rule-and-keyword escalation decides when a conversation moves from the AI to a person, and copilot mode lets any team member take over a specific conversation in one click — per conversation, per channel or for the whole workspace — with the bot pausing itself so it doesn't step on the human reply. In a multi-brand operation, that means each team works its own queue without interfering with the others.
A different AI model per brand, if needed
Since the language model is chosen per workspace (Anthropic, OpenAI, DeepSeek or OpenRouter), a marketplace can run one brand on one provider and another on a different one, depending on each team's budget or preference, without affecting the rest of the brands in the same group. It's a decision made workspace by workspace, not a global setting forced onto every store in the marketplace.
An MCP server per workspace
Each workspace also exposes its own MCP server (Model Context Protocol, the open standard created by Anthropic), so if any brand in the marketplace already uses its own internal AI for management, it can connect to its specific heysteff workspace and configure only what belongs to that brand, with no access to the others' data.
◆ the star flow
The workspace per brand with its own roles is what makes it possible to scale from one brand to many without rewriting the operation from scratch each time.