PilotR

MCP-compliant R code execution and management agent for Claude Desktop Extension.

Empower R programming with Model Context Protocol.

Overview

PilotR is an MCP (Model Context Protocol) server that allows Claude to generate, execute, and manage R scripts within a user-specified directory. It provides a complete workflow for statistical programming and data analysis through natural language interaction — no manual setup required.

Features

• Execute R code and scripts directly from Claude
• Manage R script files — create, rename, append, and write
• Set and manage working directories for analysis projects
• Inspect R objects and session state
• Read and preview output files (CSV, TSV, RDS, text, etc.)
• List and manage exported results
• Automatic workspace saving
• Sandboxed file operations for safety

Installation (No Dependencies Needed)

Step 0: Download and Set Up Claude Desktop and R

1. Download Claude Desktop and set up your Claude account. Free-tier accounts are sufficient to get started with PilotR, but the Pro accounts are recommended for more complicated tasks.
2. Download and install R runtime, install it and make sure that it is within the system path.

Step 1: Download PilotR.mcpb

Simply download the latest PilotR.mcpb file from this repository or release page.

Step 2: Install in Claude Desktop

1. Locate the downloaded PilotR.mcpb file
2. Double-click the file
3. Claude Desktop will open and prompt to install
4. Follow the on-screen instructions. (A Python runtime might need to be installed at this step. Please just follow the prompts to proceed with the installation.)

Configuration (Optional)

To streamline your workflow:

1. Open Claude Desktop
2. Go to Settings → Extensions → MCP Servers
3. Find "PilotR"
4. Enable "Allow tools to run without permission" (Auto-approve)
5. Restart Claude Desktop

Usage Examples

Once installed, you can interact with PilotR using natural language.
Here are common commands and workflows:

Common Commands

Setup & Configuration

"Set working directory to ~/my_analysis"    # Initialize workspace
"Show current state"                        # Check configuration
"Find R executable"                         # Verify R installation

File Management

"Create an R file called analysis.R"        # New script with scaffold
"Write R code to load and plot data"        # Generate and save code
"Append summary statistics to my script"    # Add to existing file
"Rename script.R to final_analysis.R"       # Rename files
"List all R files"                          # View R scripts

Code Execution

"Run my R script"                           # Execute primary file
"Run R expression: mean(1:10)"              # Quick calculations
"Execute analysis.R with arguments"         # Run with parameters
"Show R objects in workspace"               # Inspect saved data

Data Operations

"Preview data.csv"                          # View CSV/TSV data
"Read results.txt"                          # Read text files
"List all output files"                     # See generated files
"Show files created today"                  # Recent outputs

Visualization & Analysis

"Optimize my ggplot code"                   # Improve plot styling
"Check ggplot style in my script"           # Style analysis
"Create publication-ready plot"             # High-quality output

Example Workflows

Basic Data Analysis

You: Create an R script for linear regression analysis
You: Load the mtcars dataset and fit a model of mpg vs weight
You: Run the script and show the model summary
You: Create diagnostic plots for the model
You: Save the plots as high-resolution PNGs

Data Processing Pipeline

You: Set working directory to ~/data_project
You: Read my raw_data.csv file and show a preview
You: Create a script to clean missing values and outliers
You: Transform the data and save as cleaned_data.csv
You: Generate summary statistics and visualizations

Publication-Quality Visualization

You: Create a ggplot scatter plot with trend line
You: Add confidence intervals and customize colors
You: Optimize the plot code for publication
You: Save as figure1.png with 800 DPI

Exploratory Data Analysis

You: Load the iris dataset
You: Create a script for exploratory analysis
You: Generate summary statistics by species
You: Create box plots and correlation matrices
You: Run everything and show the results

Available Tools

Core Operations

• set_workdir — Set working directory for R operations
• get_state — View configuration and runtime state
• which_R — Locate R executable

File Management

• create_R_file — Create new R script
• rename_R_file — Rename existing script
• write_R_code — Write code to file (with overwrite protection)
• append_R_code — Append to existing file
• set_primary_file — Set default script for operations
• list_R_files — List R scripts in workspace

Execution

• run_R_script — Execute R script file
• run_R_expression — Evaluate R expression
• inspect_R_objects — Explore workspace objects

Data Operations

• list_exports — List exported files
• read_export — Read file contents (text or binary)
• preview_table — Preview CSV/TSV tables

Analysis Tools

• ggplot_style_check — Optimize ggplot2 code for publication

Technical Details

• Extension file: PilotR.mcpb
• Protocol: MCP (Model Context Protocol) 1.0+
• R Execution: via Rscript
• Session Persistence: .pilotr directory for state
• Security: Sandboxed file operations
• Python: Requires Python 3.12+

Acknowledgments

Special thanks to Beniamin Krupkin and Sara Smith for the inspiration of this idea and their valuable contributions.

Maintainer

Wanjun Gu ([email protected])

License

MIT