MCP (Model Context Provider)

MultitaskAI supports the Model Context Protocol (MCP), an open standard that allows applications to connect AI models with various external tools and data sources securely. Think of MCP like a universal adapter for AI – it provides a standardized way for MultitaskAI to interact with different capabilities exposed by MCP servers.

You can learn more about the protocol at modelcontextprotocol.io.

What is an MCP Server?

An MCP server is a lightweight program that exposes specific tools or data sources (like accessing a database, interacting with an API, or reading local files) through the standardized MCP interface. By connecting MultitaskAI to an MCP server, you grant the AI access to these capabilities.

Adding an MCP Server

Connecting an MCP server in MultitaskAI is straightforward:

1. Navigate to the MCP Section
   Go to the MCP section in the sidebar, or visit multitaskai.com/mcp.

2. Add a New MCP Connection
   Click the + icon or "New MCP" to start configuring a new server connection.

3. Configure the Connection

   • Name: Give your MCP connection a descriptive name (e.g., "Zapier Actions", "Local File Access").
   • Endpoint URL: Enter the URL where the MCP server is running.
   • Authentication: Currently, MultitaskAI only supports MCP servers that do not require authentication. Support for authenticated endpoints is planned for a future update.

4. Discover and Select Tools

   • Once the Endpoint URL is entered, MultitaskAI attempts to connect to the server and discover the available tools (actions) it provides.
   • Click Reload to refresh the list of tools if the server's capabilities change.
   • Use the checkboxes to select which tools from the MCP server should be available within MultitaskAI chats.
   • You can use Select All or Deselect All for convenience.

5. Save the Connection
   Click Save (or ensure it saves automatically) to store the MCP connection.

Using MCP Tools in Chats

Once an MCP connection is configured and its tools are enabled, they become available for use within the chat interface:

• Selecting Tools in Chat: In the chat input area, click the Select Tools button. A dropdown will appear, listing all available tools grouped by their MCP connection (e.g., "Zapier"). Check the boxes for the tools you want to activate for the current chat session.
• Persistence: Selected tools remain active for the duration of the chat. They are saved with the chat, so if you return later, the same tools will still be selected.
• Agent Integration: If you select an Agent that has specific MCP tools assigned to it, those tools will be automatically activated in the chat, in addition to any you may have selected manually.
• AI Interaction: When an MCP tool is active (either selected manually or via an Agent), the AI model can leverage it to perform actions or retrieve data based on your prompts. For example, if you have the gmail_create_draft tool active, asking the AI to "Draft an email..." allows it to use that specific tool.

Tips for Effective MCP Use

• Use Descriptive Names: Naming your MCP connections clearly helps you remember what tools they provide.
• Select Relevant Tools: Only enable the tools you anticipate needing to avoid cluttering the tool selection interface in chats.