MCP (Model Context Protocol)
Extend Wyatt's AI capabilities by connecting MCP-compatible tool servers.
What Is MCP?
MCP (Model Context Protocol) is an open standard that lets AI assistants connect to external tools and data sources. Think of it as a universal plug that lets Wyatt's AI talk to other services in a structured, reliable way.
When you connect an MCP server to Wyatt, the AI gains new capabilities -- like searching a specific knowledge base, pulling data from a custom API, or triggering actions in third-party software.
How Wyatt Uses MCP
Wyatt uses MCP as the backbone for many of its integrations. When you connect a service like Notion or Google Drive, Wyatt communicates with it through an MCP server behind the scenes. This architecture provides several benefits:
- Standardized communication -- Every integration uses the same protocol, which means consistent behavior and reliability.
- Tool discovery -- When an MCP server is connected, Wyatt automatically discovers what tools it provides. The AI learns what actions are available and when to use them.
- Extensibility -- Because MCP is an open protocol, you can connect servers built by the community or even create your own.
Connecting an MCP Server
- Go to Settings > Integrations > MCP Servers.
- Click Add Server.
- Enter the server URL or select from available options.
- Authenticate if required (some servers need API keys or OAuth).
- Wyatt will discover the server's available tools and make them accessible to the AI.
Once connected, the AI can use the MCP server's tools automatically during conversations. You don't need to do anything special -- just ask the AI a question and it will use the right tools to find the answer.
What MCP Servers Provide
An MCP server can expose one or more of the following:
| Capability | Description |
|---|---|
| Tools | Actions the AI can call, like "search documents" or "create a record" |
| Resources | Data the AI can read, like files, database rows, or API responses |
| Prompts | Pre-built instructions that guide the AI for specific tasks |
Tool Discovery
When you connect an MCP server, Wyatt automatically discovers all the tools it provides. These tools show up in:
- AI conversations -- The assistant can choose and invoke them when they're relevant to your request.
- Federated search -- Search tools from MCP servers contribute results to unified search.
- Automations -- MCP tools can be used as steps in your automated workflows.
You can see all available MCP tools in Settings > Integrations > MCP Servers by expanding any connected server. This is a good way to understand what capabilities a server provides.
Building Your Own MCP Server
If you have a custom data source or internal tool you'd like the AI to access, you can build your own MCP server. MCP servers can be written in any language and expose tools through a standardized API.
The key steps are:
- Implement the MCP protocol (see the MCP specification for details).
- Define the tools your server provides, including their names, descriptions, and input schemas.
- Host your server at a URL that Wyatt can reach.
- Connect it in Settings > Integrations > MCP Servers.
Building an MCP server requires some technical knowledge. If you're not a developer, you can still benefit from MCP by connecting pre-built servers from the community.
Related
- Integrations overview -- all available integrations
- Connections -- setting up OAuth-based service connections
- Federated Search -- how connected services contribute to search