The multi-tenant control plane
alice_hub: the UI and orchestration layer that owns users, companies, auth, agent CRUD, chat threads, dashboards, and audit logs — and routes to the right runtime.
The stack's front door
Hub is the multi-tenant control plane that sits in front of your A.L.I.C.E. runtime infrastructure. It's the UI your team uses to manage agents, monitor executions, audit activity, and configure company settings. It's also the routing layer that dispatches incoming requests to the right runtime.
Built as a Paperclip fork on a pnpm monorepo — Express + Drizzle + ws server on the backend, Vite + React 19 + assistant-ui on the frontend. The internal WebSocket at /api/companies{id}/events/ws is UI↔️Hub-server only. Agent dispatch goes over OpenAI-compatible HTTP to the runtime.
Hub owns the framework resolver — given a request, it determines which runtime should handle it (alice-runtime, OpenClaw, or another) and forwards the call. Your agents and your users don't need to know which runtime is running underneath.
Paperclip fork architecture
Backend: Express + Drizzle ORM + WebSocket server
Frontend: Vite + React 19 + assistant-ui
Auth: Better-Auth (SSO, MFA, session management)
Monorepo: pnpm workspaces
Everything Hub manages
Dashboard
Real-time agent activity, system health, and execution traces at a glance
Multi-Tenant
Users, companies, and access control — scoped per organization
Better-Auth
Enterprise-grade authentication with SSO, MFA, and audit logging
Framework Resolver
Routes to alice-runtime, OpenClaw, or any OpenAI-compatible runtime
Audit Log
Every agent action, tool call, and decision — fully traceable
Role-Based Access
Fine-grained permissions for admins, operators, and end users
Hub routes the full stack
Hub → picks a framework → calls Runtime over OpenAI-compat HTTP → Runtime loads the Agents workspace + persona → tool-calls Memory via FFI.
Ready to take control?
Hub is the operational layer that makes the full A.L.I.C.E. stack usable by a real team.