MCP (Model Context Protocol)AI and LLM Providers
Your Agents Reach Any System That Publishes an MCP Server
The Model Context Protocol (MCP) is an open standard for exposing a system's capabilities as tools a model can call. Vendors are publishing MCP servers for their own products. AI Intelligence Hub connects to those servers as a client, discovers their tools and lets an agent or a workflow call them with configuration only. The server and the system behind it stay as they are, and so do their records.
What you can do together
Reach systems this catalog does not name
AI Intelligence Hub can call any system that publishes an MCP server, with configuration only.
Let an agent choose the tool
An AI Intelligence Hub agent gets a server's tools beside its own, such as library search, reading content and summarizing, and decides which to call.
Run the same call every time
An MCP node in an AI Intelligence Hub workflow calls one named tool with fixed arguments on every run.
Put a person before the call
An approval node in an AI Intelligence Hub workflow holds a call that would change a record until someone approves it.
Use new tools without rework
When a server adds a tool, AI Intelligence Hub can call it from the next bind.
How it connects
AI Intelligence Hub is the MCP client. A server is configured with a name, its URL, the transport and optional HTTP headers, which carry the authentication token. Streamable HTTP is the recommended transport for a web-hosted server. Server-Sent Events is carried for servers on the earlier MCP transport, and WebSocket for those that publish a socket endpoint. All three are HTTP-based and suit a server reached over a network. A local stdio server, which runs as a child process on the machine that starts it, is not among them.
The tool list is fetched live when the connection binds, so a tool the server gains is usable without reconfiguration. The MCP node runs one named tool in a workflow, with arguments fixed at design time. The MCP tool hands the server's tool list to an agent's model, which chooses a tool and supplies the arguments at run time.
A call sends a tool's name and arguments, and the result returns to the workflow or the agent. The direction is client only. The platform does not publish itself as an MCP server, so systems calling in use the REST API or webhooks.
MCP (Model Context Protocol)AI and LLM Providers
A scenario
- SetupThe IT lead adds the server to AI Intelligence Hub with a name, its URL, Streamable HTTP and a token in a header. The Maintenance Desk agent gets its tools beside library search and read content.
- 06:40A technician asks why press line 3 keeps faulting and how it was fixed last time. The agent calls the work order tool for the line's history, then reads the transcript of the repair procedure video.
- 06:42He sees one answer listing the recent work orders and walking through the procedure. He asks for a short checklist, and the agent summarizes the steps.
- 09:15A new work order opens, and the maintenance system calls a workflow's HTTP trigger. An MCP node fetches the order, a library search finds the procedure, and a language model node drafts a note. The shift supervisor approves it at an approval node, and a second MCP node posts it to the order.
- LaterThe vendor adds a spare parts tool to its server. The agent can call it from the next bind, with no configuration change.
What stays where
The MCP server and the system behind it stay yours
The server runs wherever you or its vendor host it, and the records stay in the system behind it. AI Intelligence Hub calls only the tools the server exposes.
AI Intelligence Hub sends tool calls and receives results
A tool's name and arguments go out and the result comes back. The token travels in the header you configure, so what the connection can do is what that token is granted.
Nothing is replaced
Connecting a server is configuration in AI Intelligence Hub, and nothing is installed on the system behind it.
Where processing runs
The platform runs as shared or dedicated SaaS, in your own cloud, on your own servers or air-gapped. Each tool runs on its MCP server, which must be reachable over the deployment's network. An air-gapped deployment reaches servers inside its own network and self-hosts its language models through vLLM or Ollama.
Products and solutions
Next step
See it on your own MCP (Model Context Protocol) instance.
We will show the connection made, the data moving and the output, then size it for your deployment.
Contact VIDIZMO
sales@vidizmo.ai
+1 571-969-2180
vidizmo.ai