The MCP ecosystem has been overwhelmingly JavaScript and Python. io.github.IMNMV/clauder opens the door to R.
The Server
ClaudeR is an R package that bridges RStudio to MCP-compatible AI assistants — Claude Desktop, Claude Code, Codex, Gemini CLI, and Cursor. Install the R package, run uvx clauder-mcp, and the AI can execute code in your live RStudio session and see the results in real-time. The README describes it as:
"ClaudeR is an R package that forges a direct link between RStudio and MCP configured AI assistants like Claude Code or Codex. This enables interactive coding sessions where the AI can execute code in your active RStudio environment and see the results in real-time."
— ClaudeR README
The feature set goes well beyond basic code execution. Fifteen MCP tools cover: synchronous and asynchronous R execution (long-running jobs via callr), plot capture, interactive widget reading (plotly, DT, leaflet — with pagination for large outputs), active document reading and editing, file access, session history, and task tracking. Security restrictions block dangerous functions like system(), unlink(), and file.remove() by default.
Multi-Agent Orchestration
The standout feature is multi-agent support. Multiple AI agents can connect to the same R session simultaneously — or spread across separate RStudio windows. Each agent gets a unique ID on startup, and console output, log files, and execution history are all attributed per agent. From the README:
"On its very first tool call, each agent receives a context briefing with its own ID, any other agents active on the session, and the log file path, giving it full awareness of the shared environment without any manual setup."
— ClaudeR README
Session discovery works through files written to ~/.claude_r_sessions/ — agents find sessions automatically with no hardcoded ports or manual configuration. One agent could be cleaning data while another builds a model on the same session, each aware of the other's presence.
Why It Matters
R remains the language of choice for a large segment of statisticians, bioinformaticians, epidemiologists, and academic researchers. These are people who work in RStudio daily but have had no natural path into the MCP ecosystem. ClaudeR changes that. The combination of live session access, plot capture, and htmlwidget support means an AI agent can do what a human collaborator would — run code, look at the chart, iterate. The demo video shows this in action.
The repo has been in development since March 2025 — nearly a year of commits, MIT-licensed, published on PyPI as clauder-mcp. 128 stars and 14 forks. Score: 72. No flags. The highest-scoring new entry in this batch after PEAC Protocol.
Sources: IMNMV — GitHub · ClaudeR — repo · demo video · PyPI · Scorecard: io.github.IMNMV (score 72)