r-mcptools

mcptools: R as Model Context Protocol Server

Overview

mcptools lets AI agents INTO your R session. Agents (Claude Code, VS Code, etc.) can run R code, query data, call functions in your live R environment.

btw vs mcptools: btw gives tools TO ellmer (R → LLM). mcptools lets agents INTO R (LLM → R).

Install: install.packages("mcptools")

References

Read references/API.md before writing code.

• references/API.md - Complete function reference
• references/package-docs.md - Setup and configuration
• references/server.md - MCP server setup and agent configuration

When to Use

• Let Claude Code/VS Code run R code
• Give agents access to live R session
• Expose R functions as MCP tools
• Query R data stores from agents

When NOT to Use

• Want to call LLMs FROM R (use r-ellmer)
• Just need R object descriptions (use r-btw)
• Building chatbot in R (use r-ellmer + r-btw)

Quick Reference

library(mcptools)

# In R console (each startup, NOT in scripts)
mcp_session()

# Now Claude Code can:
# - Run R code
# - Query data frames
# - Call your R functions
# - Access documentation

CRITICAL: Call mcp_session() in console at each R startup, NOT in scripts.

Common Mistakes

Issue	Solution
btw vs mcptools confusion	btw = tools TO chat; mcptools = agent INTO R
mcp_session() in scripts	Call in console, each startup
Expecting clipboard copy	mcptools is for agent access, not clipboard
Session not detected	Restart R, call mcp_session() in console

Core Functions

Session Management:

• mcp_session(): Initialize MCP session (console only, per startup)
• mcp_mcp_server(): Launch MCP server

Advanced

See references/ for:

• API.md: Complete function reference
• server.md: R as MCP server details
• Package docs: Full package documentation

Integration

With Claude Code: Call mcp_session() in R console With ragnar: Agents can query ragnar stores Cross-package patterns: See r-ai meta-skill