rmcp

<!-- README.md is generated from README.Rmd. Please edit that file --> # rmcp <!-- badges: start --> [](https://lifecycle.r-lib.org/articles/stages.html#experimental) [](https://CRAN.R-project.org/package=rmcp) [](https://github.com/simonpcouch/rmcp/actions/workflows/R-CMD-check.yaml) <!-- badges: end --> The goal of rmcp is to enable LLM-enabled tools like Claude Code to learn about the R packages you have installed using the [Model Context Protocol](https://modelcontextprotocol.io/) (MCP). Tools configured with rmcp can peruse package documentation and learn to use R packages even if they weren’t included in their training data. rmcp is written primarily in TypeScript but is distributed as an R package for ease of distribution/install for R users. Installing the package ensures you have the needed dependencies and provides a shortcut, `mcp_config()`, to help you configure the tool with other applications. > IMPORTANT: This is an early proof of concept. Use at your own risk! ## Installation You can install the development version of rmcp like so: pak::pak("simonpcouch/rmcp") rmcp can be hooked up to any application that supports MCP. Use `mcp_config()` to generate the config for your machine for common applications. For example, **Claude Code**: library(rmcp) mcp_config("Claude Code") #> In a terminal, run: #> #> claude mcp add -s "user" r-mcp node /Users/simoncouch/Library/R/arm64/4.4/library/rmcp/node/dist/index.js Or, **Claude Desktop**: mcp_config("Claude Desktop") #> Run `file.edit(~/Library/Application Support/Claude/claude_desktop_config.json)` #> #> Then, paste the following: #> { #> "mcpServers": { #> "r-mcp": { #> "command": "node", #> "args": ["/Users/simoncouch/Library/R/arm64/4.4/library/rmcp/node/dist/index.js"] #> } #> } #> } ## Example In Claude Desktop, I’ll write the following: > Using the R packages I have installed, write code to download data on > flights in/out of Chicago airports in 2024. In a typical chat interface, I’d be wary of two failure points here: 1. The model doesn’t know which packages I have installed. 2. If the model correctly guesses which packages I have installed, there may not be enough information about how to *use* the packages baked into its weights to write correct code. <img src="https://github.com/user-attachments/assets/821ea3d6-4e30-46d6-ab9b-301276af2c35" alt="A screencast of a chat with Claude. I ask 'Using the R packages I have installed, write code to download data on flights in/out of Chicago airports in 2024.' and, after searching through the documentation of my currently installed R packages, Claude writes R code to do so." width="100%" /> Through first searching through my installed packages, Claude can locate the anyflights package, which seems like a reasonable solution. The model then discovers the package’s `anyflights()` function and reads its documentation, and can pattern-match from there to write the correct code.