Getting Started with mcpr

library(mcpr)

Introduction

The mcpr package allows you to create Model Context Protocol (MCP) servers in R. This vignette will guide you through building a very simple MCP server with a basic addition tool.

Creating a Simple MCP Server

Let’s create a simple MCP server that can add two numbers:

# Create a new MCP server
mcp_server <- new_server(
  name = "Simple Adder",
  description = "A basic MCP server that adds two numbers",
  version = "1.0.0"
)

# Create a simple addition tool
add_tool <- new_tool(
  name = "add",
  description = "Adds two numbers together",
  input_schema = schema(
    properties = list(
      a = property_number("First number", "First number to add", required = TRUE),
      b = property_number("Second number", "Second number to add", required = TRUE)
    )
  ),
  handler = function(input) {
    result <- input$a + input$b
    response_text(paste("The sum is:", result))
  }
)

# Add the tool to the server
mcp_server <- add_capability(mcp_server, add_tool)

Using Standard Input/Output

The simplest way to interact with your MCP server is through standard input/output using the serve_io function:

# Start the server using standard input/output
serve_io(mcp_server)

When you run serve_io(mcp_server), your R session will wait for input in JSON-RPC 2.0 format. You can then communicate with your server by sending JSON-RPC requests.

Testing the Server

To test your server, you can run the above code in an R script and interact with it using the command line. For example:

  1. Save the code in a file named mcp_adder.R
  2. Run it from the command line: Rscript mcp_adder.R
  3. Send a JSON-RPC request to standard input:
{"jsonrpc": "2.0", "id": 123, "method": "tools/call", "params": {"name": "add", "arguments": {"a": 15, "b": 27}}}

The server will respond with:

{"jsonrpc": "2.0", "id": 123, "result": {"content": [{"type": "text", "text": "The sum is: 42"}]}}

Next Steps

Once you’re comfortable with the basics, you can:

  1. Add more complex tools with different operations
  2. Create resources to serve data
  3. Implement prompts for text generation

See the other vignettes and function documentation for more details on these advanced features.