Use MCP servers in prompts

Connect to Model Context Protocol servers to access external tools and data sources from your prompts

The Model Context Protocol (MCP) is an open protocol that enables AI models to securely access external data sources, tools, and resources through a standardized interface. Adaline’s Editor lets you connect to remote MCP servers and use their tools directly in your prompts — enabling powerful integrations without custom backend code.

Enable MCP

1. Select a compatible model

   Open the model selector and choose an LLM that supports MCP integration.

   MCP integration currently requires selecting an Anthropic model. If you need MCP support for other providers, contact [e-mail] for a private preview.

2. Open MCP settings

   In the model settings panel, select MCP.

   
3. Toggle MCP on

   Set the MCP toggle to On to enable the integration.

   

Once enabled, the Editor displays the MCP configuration panel where you can add your server settings:

Configure an MCP server

Each MCP server requires a JSON configuration object. Here is the complete schema:

{
  "type": "url",
  "url": "https://example-server.modelcontextprotocol.io/sse",
  "name": "example-mcp",
  "tool_configuration": {
    "enabled": true,
    "allowed_tools": ["example_tool_1", "example_tool_2"]
  },
  "authorization_token": "YOUR_TOKEN"
}

Configuration reference

Field	Type	Required	Description
type	String	Yes	Currently only "url" is supported.
url	String	Yes	The MCP server URL. Must start with https://.
name	String	Yes	A unique identifier for the server. Choose a descriptive name.
tool_configuration	Object	No	Controls which tools from the server are available.
tool_configuration.enabled	Boolean	No	Whether to enable tools from the server (default: true).
tool_configuration.allowed_tools	Array	No	Restrict which tools are allowed. By default, all tools are available.
authorization_token	String	No	OAuth authorization token, if required by the server.

Connect to multiple servers

You can connect to multiple MCP servers simultaneously by providing an array of server configurations:

[
  {
    "type": "url",
    "url": "https://mcp.example1.com/sse",
    "name": "mcp-server-1",
    "authorization_token": "YOUR_TOKEN_1"
  },
  {
    "type": "url",
    "url": "https://mcp.example2.com/sse",
    "name": "mcp-server-2",
    "authorization_token": "YOUR_TOKEN_2"
  }
]

Each server’s tools become available to the model independently. The model can choose which server’s tools to invoke based on the conversation context.

Authentication

Many MCP servers require OAuth authentication. To obtain an access token:

1. Use the MCP Inspector or your server provider’s OAuth flow to generate a token.
2. Paste the token into the authorization_token field in your server configuration.

Keep your authorization tokens secure. If your MCP server’s tokens have an expiry, you will need to refresh them periodically.

Next steps

Use Tools in Prompts

Define custom tools with JSON schemas and HTTP backends.

Tool Calls in Playground

Test MCP-powered and tool-powered prompts in the Playground.