AI Studio

Status of AI Studio — what it is intended to be, that it is a preview in Cockpit with no public API, and the AI-adjacent things that do work today.

AI Studio is the intended home for the platform's AI: assistants that answer grounded questions about a tenant's own commerce data, findings surfaced without being asked, and — further out — actions taken inside rules an operator sets. It is meant to work on real products, orders, customers and pricing rather than sit beside them as a chatbot.

AI Studio is a preview, and there is no public API. Every AI Studio screen in Cockpit is badged as a demo and renders fixture data; its backend is not wired up. So agents, knowledge, tools, insights, autonomy rules, traces and credits are not usable by a customer today — in Cockpit or from code, and there are no AI Studio endpoints on the gateway at https://api.revenexx.com.That is a statement about AI Studio, not about AI on the platform. Some AI capability is real and usable now; it is listed below, and none of it is AI Studio.

This page exists so you can plan around that honestly. Don't build against AI Studio yet, and don't quote its Cockpit screens to a customer as shipped functionality.

What does work today

Several AI-adjacent things are real and usable now. None of them is AI Studio.

A translation node in Integration Studio

The built-in node catalogue includes a machine-translation node. Drop it into a workflow and it translates field values as part of a run — useful for localising catalogue copy on import without writing any code or holding a translation API key in your own service. Credentials for the translation provider are configured once on the tenant and referenced by the node, the same as any other node credential.

See Integration Studio for workflows, and Node types for the catalogue it sits in.

MCP servers an AI client can drive

If your goal is "let an AI assistant work with revenexx", the Model Context Protocol is the path that exists today, and it doesn't depend on AI Studio at all.

  • This documentation, as an MCP server. Hosted, public, no credentials: https://revenexx.dev/mcp. An MCP-capable assistant can search and read these docs while it writes code against the platform. Setup per client is covered under MCP.
  • The Integration Studio workflow MCP server. Integration Studio exposes its workflow management over MCP at https://integrations.revenexx.com/mcp/workflows, authenticated with the token of a signed-in person and acting on one tenant. An assistant connected to it can list, create, update and delete workflows, add, replace and delete their triggers, activate and deactivate them, lay out their nodes, and start runs — as well as inspect runs, credentials and state. Nine of its 22 tools are write operations and there is no read-only mode, so the token you give an assistant is the whole boundary. Read What an AI agent can change before you connect one for a customer.
  • api-mcp, the revenexx API as tools. Four generic tools that discover and execute operations from your tenant's API contract, only one of which writes, refused by default. It is reached through AI access arranged with revenexx, not at a URL you configure, so it is described on What an AI agent can change rather than set up here.

Where to go next

  • MCP — connect an AI client to the documentation, with per-client setup.
  • Integration Studio — workflows, nodes and triggers.
  • Events — the activity stream to feed your own AI pipeline.
  • Changelog — where an AI Studio API would be announced when there is one.
Was this page helpful?