Skip to main content
Your product and OpenGeni stay separate systems. Your product owns its users, tenants, business data, and pages. OpenGeni owns agent sessions, turns, event history, approvals, files, and compute. They meet at one backend route and at the tools you expose.
  1. The browser runs the stock SDK client against your own origin, /api/opengeni.
  2. Your proxy authenticates each request with your existing session check, resolves the user’s workspace, and forwards only the routes the conversation needs, acting as that user.
  3. OpenGeni runs the agent. When it needs product data, it calls your MCP server or OpenAPI Integration with a credential you issued for that user.
The organization API key stays on your server. The browser can never pick another workspace, choose tools, or send credentials.

Mapping your product onto OpenGeni

A workspace is the sharing boundary for sessions, files, knowledge, connections, and integrations. Use one per customer when their users share those; see Users, tenants & privacy for other choices.

Who does what

Link records by opaque ids. Keep your business data in your product and let the agent fetch it through tools.
Using a coding agent? The opengeni-client skill covers this.