mcptools (R)

Posit’s CRAN package that exposes R to Claude over MCP — either as a standalone server that runs R code, or as a bridge into an already-running interactive R session with your data loaded.


Type	MCP server (R package; also an MCP client for R via ellmer)
Supplier	Posit Software, PBC — maintainer Simon Couch
Availability	GA — CRAN 1.0.1, published 2026-07-27
Pricing	Free / OSS — MIT
Capabilities	Read/Write — runs R in your session; the exact surface is whatever tools you register
Verified	works · 2026-08-17
Security	cleared · 2026-08-17 — MIT confirmed via raw LICENSE.md, provenance matches Posit, no advisories

How to install

Install the package from an R console:
```
install.packages("mcptools")
```
R 4.1.0 or newer.

Claude Code — register the server:

claude mcp add -s "user" r-mcptools -- Rscript -e "mcptools::mcp_server()"

Claude Desktop — add to claude_desktop_config.json:
```
{
  "mcpServers": {
    "r-mcptools": {
      "command": "Rscript",
      "args": ["-e", "mcptools::mcp_server()"]
    }
  }
}
```
Fully quit and relaunch Claude Desktop after editing. (Unverified — upstream documents only the claude mcp add form; this JSON is the direct equivalent of that command but is not shown in the package docs. On Windows, Rscript must be on PATH or given as an absolute path.)
Optional — connect to your live R session. Add this to your .Rprofile so interactive sessions announce themselves to the server:
```
if (interactive() && requireNamespace("mcptools", quietly = TRUE)) {
  mcptools::mcp_session()
}
```
The server then picks the session whose working directory matches its own, so a Claude Code instance launched inside a project connects to that project’s R session automatically.

mcp_server() is launched by Claude over stdio — you do not run it yourself in a terminal.

What it does

Out of the box mcp_server() exposes only session-plumbing tools — list_r_sessions() and select_r_session() — and deliberately nothing that touches your data. Useful capability comes from passing your own tools to the tools argument: any function wrapped with ellmer::tool() can be registered, including a general run_r_code() tool if you want the agent to execute arbitrary R.

Because it runs in a real R session, this is the practical route to driving Bioconductor from Claude — DESeq2, SummarizedExperiment, limma, flowCore and the rest have no MCP servers of their own.

Primary use cases: driving Bioconductor and R statistics workflows from Claude, letting Claude inspect objects in a live analysis session, exposing lab-specific R functions as agent tools.

Notes

Registering a run_r_code()-style tool gives the model arbitrary code execution in your R session, with whatever filesystem and credential access that session has — the package docs frame this as an option “for the brave” rather than a default. Prefer narrowly-scoped ellmer::tool() wrappers for anything shared or automated.

mcptools is also an MCP client: R code using ellmer can pull tools from third-party MCP servers into a chat, which is the reverse direction from the server use described above.

Several catalogued R-based skills — DESeq2 differential expression, cytometry QC, gating analysis, clustering and phenotyping — assume an R runtime is reachable. This package is the general-purpose way to provide one.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.