How X-MSP serves every client from one console without mixing their data
Managed service providers need one platform for dozens of customers, and AI that never puts one client's knowledge into another client's reply. Here is how X-MSP keeps them apart.
A managed service provider lives with a contradiction. Operators want one console for every client. Each client wants to be sure no one else can see its tickets, knowledge base or users. Once AI drafts the replies, the stakes go up: one wrong search result can put a client's internal fix into another client's inbox.
X-MSP is X-ITSM built for exactly this.
One platform, strict client boundaries
- Multi-tenant by design. One database with strict client isolation. Operators switch tenants without logging in again.
- Role-aware access. Separate MSP and client realms, with row-level filters built into every query.
- Branded client portals. A subdomain, logo and theme for each client. End users see only their own data.
- Per-client SLAs and catalogs. Each customer has its own SLAs, priorities and service catalog.
A new client in seconds
Clone-on-onboard creates a new client from your template tenant, fully set up, in seconds. Cross-client reporting gives operators one view of tickets, SLAs and load across every tenant.
AI that respects tenant boundaries
Every AI feature limits its search, prompts and output to the client the ticket belongs to.
- Agent Assist summarises long threads, drafts a reply in the client's preferred language and suggests fixes from that client's similar resolved tickets.
- Tenant-safe search. Every knowledge article is embedded and tagged with its client. Searches run only inside the client of the request, and this is enforced where the query is built.
- A multi-channel chatbot on Feishu, WeChat, WeCom, Youdu and the web portal classifies, searches, responds and escalates. It answers from each client's own knowledge base, escalates when the mood turns, and gives the MSP operator a supervisor view.
The operator sees everything. The AI sees only what the current ticket is allowed to see.
A model per client, if you need one
Choose OpenAI for one client, Anthropic for another and a local Ollama model for a regulated one. The embedding provider is set separately. Point the whole stack at a local llama.cpp or vLLM server to run it air-gapped. A client's data reaches a public model only if you choose it.
Try X-MSP
Spin up a demo in minutes, and bring your own data when you're ready. See X-MSP or talk to us.
