ClaudeR connects RStudio to various Large Language Model (LLM) agents, including Claude Code, Codex, Gemini, and Qwen Code, via the Model Context Protocol (MCP). It empowers R users, researchers, and data scientists by enabling interactive coding, multi-agent orchestration, automated manuscript auditing, and data annotation directly within their R development environment. This integration allows LLMs to execute R code, view generated plots in real-time, and assist with complex analytical tasks, significantly enhancing productivity and research rigor.

How It Works

ClaudeR leverages the Model Context Protocol (MCP) to establish a secure, bidirectional communication channel between LLM agents and an R add-in running in RStudio. A Python MCP server acts as a bridge, forwarding AI requests to RStudio for execution. The R session then returns results, plots, or errors back to the AI. This architecture ensures that LLMs can interact with the R environment safely through well-defined interfaces, enabling features like direct code execution, environment variable access, and asynchronous processing without compromising user control.

Quick Start & Requirements

• Primary Install: devtools::install_github("IMNMV/ClaudeR") in RStudio.
• AI Setup: Use install_clauder() for desktop apps (Claude Desktop, Cursor) or install_cli(tools = "...") for CLI agents (Claude Code, Codex, Qwen Code, Gemini). Defaults to uvx for zero-config Python dependency management.
• Start Server: Run claudeAddin() in RStudio to launch the UI in the Viewer pane and click "Start Server".
• Prerequisites: R, RStudio, devtools package. Python is managed by uvx by default.
• Docs: See llms-install.md for detailed automated setup instructions.

Highlighted Details

• Reviewer Zero: A 4-pass protocol for automated academic manuscript auditing, verifying statistical claims against R code and checking references via CrossRef.
• AI-Driven Data Annotation: Tools for row-by-row CSV labeling, supporting interactive sessions or isolated subprocesses per row for consistency or privacy.
• Multi-Agent Coordination: A built-in protocol enabling multiple AI agents to collaborate within a single R session, negotiating tasks and cross-checking outputs.
• R Best Practices Protocol: A guided workflow for reproducible statistical analysis, covering EDA, model building, diagnostics, and reporting.
• read_file Tool: Allows agents to read any text file from disk, facilitating session continuity and analysis of external scripts or logs.
• Async Execution: execute_r_async enables long-running R code (simulations, model fitting) in a separate process, keeping the main R session responsive.

Maintenance & Community

The README does not provide specific details on maintainers, community channels (like Discord/Slack), or project roadmap beyond feature updates.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Permissive MIT license generally allows for commercial use and integration into closed-source projects.

Limitations & Caveats

ClaudeR operates as a "supervised power tool," executing code within the user's live R session; users must review AI-generated code. While system commands and file deletion are blocked, agents can still read files, install packages, and consume resources. Each R session can only connect to one desktop AI application at a time, though multiple CLI agents can coexist. Asynchronous jobs run in separate processes, requiring explicit data serialization for environment sharing. Plot display may necessitate specific functions or print() calls.