MCP Client User Guide
These guides cover talking to a language model, connecting to MCP servers and routing tool calls between the two, and the dialogs the component ships for configuring all of it. Start with Getting started if you do not have a client answering questions yet.
Guides
| Guide | Scope | API starting points |
|---|---|---|
| Querying a language model | Choosing a service and its model, supplying API keys safely, generation settings and the system role, the tool collection the model sees, sending a query, and interpreting the result. | TTMSMCPClient, TTMSMCPCloudAI, TTMSMCPCloudAISettings |
| Connecting to MCP servers | The server collection, STDIO, SSE and streamable HTTP transports and where each one runs, JSON configuration files, tool discovery and routing, approving tool calls, elicitation, and logging. | TTMSMCPClientServers, TTMSMCPClientServerItem, TTMSMCPClientCapabilities |
| The settings and tools dialogs | Assigning a client to the settings dialog, persisting the keys and the server list it edits, and showing the tools dialog for one connection. | TTMSMCPClientSettingsDialog, TTMSMCPClientToolsDialog |
Related components
- MCP Server is the other end of the protocol. A server you write with it is reachable from this client over any of the three client transports.
- Cloud AI is the component behind the
LLMproperty. Everything it offers — file context, assistants, speech, transcription — is reachable throughLLMon a client. - Tools describes what a server exposes and this client discovers.
- MCP Transports covers both ends of each pipe.