ModelContextProtocol JL

Open a Julia REPL and install the package from the Julia general registry using the built-in package manager.

using Pkg
Pkg.add("ModelContextProtocol")

Create tool definitions that expose Julia functions to the AI model. Each MCPTool requires a name, description, parameter definitions with types, and a handler function that returns TextContent.

using ModelContextProtocol, Dates, JSON3

time_tool = MCPTool(
    name = "get_time",
    description = "Get current time in specified format",
    parameters = [
        ToolParameter(
            name = "format",
            type = "string",
            description = "DateTime format string (e.g. \"yyyy-mm-dd\")",
            required = true
        )
    ],
    handler = params -> TextContent(
        text = JSON3.write(Dict("time" => Dates.format(now(), params["format"])))
    )
)

Instantiate the server with your tools and start it. By default it uses stdio transport, which is what Claude Desktop expects. The server blocks until interrupted.

server = mcp_server(
    name = "my-julia-server",
    version = "1.0.0",
    tools = [time_tool]
)
start!(server)

Save your server code to a .jl file so it can be launched by Claude Desktop as an external process. Name it something like server.jl in a dedicated directory.

# Save as ~/mcp-julia/server.jl
# Then verify it runs:
julia ~/mcp-julia/server.jl

Add an entry to claude_desktop_config.json that launches your Julia script. Claude Desktop will start the process and communicate over stdio.

{
  "mcpServers": {
    "julia-server": {
      "command": "julia",
      "args": ["/Users/yourname/mcp-julia/server.jl"]
    }
  }
}

Restart Claude Desktop after saving the config. The Julia server will start automatically. Ask Claude to use one of the tools you defined, such as getting the current time.