ClaudeR MCP
MCP server that connects Claude Code and Codex to an open R or Python session in RStudio and Positron, where your data lives
- 359 ★ on the repository
- Last commit
Broad access
Why this level
- The agent runs arbitrary R in your session, reads any file you can read and reaches the network
- Third-party manuscripts, CSV files and HTML widgets may hide instructions aimed at the model, and the code runs at once
- The command guard blocks only obvious calls and can be bypassed
Install
- Works in
Manual install
devtools::install_github("IMNMV/ClaudeR")Run it in the R console. Requires R 4.0 or newer and devtools installed.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent runs arbitrary R in your session, reads any file you can read and reaches the network
- Third-party manuscripts, CSV files and HTML widgets may hide instructions aimed at the model, and the code runs at once
- The command guard blocks only obvious calls and can be bypassed
This is third-party code. Review the repository files before installing.
What it does
ClaudeR connects an AI agent to the R session you are already working in. The agent runs code where your cleaned data and fitted models live, reads objects in the environment, sees the plots it draws and edits the file open in your editor. A companion Python package serves an open Python session, and both work in RStudio and Positron. Setup is an R package plus the clauder-mcp Python bridge, which runs through uvx. The server starts from RStudio with claudeAddin(), and install_clauder() for Claude Desktop and Cursor or install_cli() for Claude Code, Codex, Qwen Code and Antigravity CLI connects the client. Several agents attach to one session and share the work, each with its own name, history and code log. You can open separate sessions so agents do not interfere. Long code runs in the background and is polled by job ID. The project includes modes for researchers. Reviewer Zero reads a manuscript block by block and checks every number against what your code produced, writing findings as Word comments. Other modes verify references through CrossRef and OpenAlex, select abstracts for a systematic review with Cohen's kappa, and turn a session log into a Quarto notebook. Checkpoints save the environment before a risky step, and a rollback works even after R restarts.
Who it is for. For researchers, analysts and statisticians on R and Python who want an agent to work inside their own session and check a manuscript against the code.
Good fit when
- Data is already loaded and cleaned in an RStudio session and the agent should work with it without reloading files
- You need to check a manuscript's numbers, p-values and confidence intervals against what your R code produced
- Before submission you want references verified against CrossRef, with retraction flags
- Two agents, for example Claude Code and Codex, share one analysis and exchange messages through a common log
- The agent is about to delete or overwrite objects and you want a checkpoint with rollback first
Not a fit when
- You do not work in RStudio or Positron: the server is built into the environment as an add-in, and without it the agent has nothing to connect to
- You want the agent on an isolated server with no file access: ClaudeR has no sandbox
- Async code must see variables of the main session: the background process does not share them
Example request
Run the regression on the survey data frame in my session, plot the residuals and check the results against Table 2 in manuscript.docxMCP
- Transport
- stdio
- Authentication
- not required
Limitations
- The agent runs arbitrary R in your open session. validate_code_security() rejects system(), shell() and recursive deletes, but it is a regex blocklist rather than a sandbox and is easy to bypass.
- 127.0.0.1 does not protect against other processes. Requests with an Origin header are always blocked, while the token check is a Require auth token checkbox that is off by default and needs clauder-mcp 0.6.0 or newer.
- One R session connects to one Claude Desktop or Cursor app, while several CLI agents can share it.
- Async jobs run in a separate process through callr without the main environment.
- Third-party documents, CSV files and HTML widgets count as untrusted input on a path to code execution, so work in a project folder and read the log.
How to disable. Click Stop Server in the add-in panel or stop it from the R console. Then remove the clauder server entry from the client config: claude mcp remove clauder for Claude Code, or delete the block in the MCP settings file for Claude Desktop or Cursor. remove.packages("ClaudeR") removes the R package. Session logs stay in the project folder; delete them if needed.
FAQ
How do I connect RStudio to Claude Code?
In R run devtools::install_github("IMNMV/ClaudeR"), then library(ClaudeR) and install_cli(tools="claude"). Then start claudeAddin() in RStudio and click Start Server. For Codex use tools="codex", and for Claude Desktop and Cursor use install_clauder().
Do I need to install Python?
The bridge runs through uvx, so you only need the uv manager. If uvx is unavailable, the installers accept use_uvx=FALSE and a Python interpreter path in python_path.
What does Reviewer Zero do?
It reads the manuscript in blocks, records every claim in a registry, verifies the extraction and recomputes values from your code. reconcile_values walks all the numbers in the document and matches them to what the code produced. Findings can be written as Word comments in a copy of the .docx file.
Can the agent damage my data?
It can: it reads any file you can read, overwrites objects and reaches the network. Call checkpoint_session before a risky step, and restore_session rolls back. Logging is on by default and records every line and the agent that ran it.
Does ClaudeR work with Python and Positron?
Yes. The companion clauder-py package connects an open Python session, and both modes work in RStudio and Positron.
Discussion
No questions or reviews yet.
Related
MCP serversA graph database with git-style branches: the server exposes it to agents over MCP at /mcp, and a skill teaches the agent schemas and queries
Connects a Zotero library to Claude and other agents: search papers, annotate, cite, and search by meaning
A desktop SQL workspace with a built-in MCP server: the agent reads the schema and runs queries in the same client you already use
A Claude Code plugin and MCP server: over 1000 scientific tools for biology, chemistry and medicine, data-check commands and routing skills
MCP server and skills for product analytics: user paths, funnels, segments and transition graphs over event logs
A lightweight database workspace with an MCP server: the agent reads the schema, runs queries and explores data in PostgreSQL, MySQL and more
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| ClaudeR (this entry) | no | not required | 359 | |
| Academic Research Skills | no | 54,527 | a skill set for research and review that does not attach to a open R session | |
| PaperOrchestra | no | 694 | a multi-agent paper-writing pipeline without running code in your session |
README in short
The project documents more than forty tools and separately covers its security model with a warning about instructions injected through documents. The author keeps recent changes in a changelog right on the main page.