IMNMV/ClaudeR

What it solves

ClaudeR connects AI agents (like Claude Code or Codex) directly to a live R or Python session in RStudio or Positron. This allows agents to perform interactive data analysis, manage cleaned data and fitted models in real-time, and automate the auditing of research manuscripts without the user having to manually copy-paste code or data between the AI and the IDE.

How it works

It uses the Model Context Protocol (MCP) to create a bridge between the LLM agent and the active coding session. The agent can execute code, read environment objects, view plots, and edit files directly in the editor. It supports multi-agent orchestration where several agents can share a single session or run in separate sessions to work on different datasets independently.

Who it’s for

Researchers and data scientists who use R or Python and want to integrate LLM agents into their analysis workflow for tasks like data cleaning, model fitting, and manuscript preparation.

Highlights

• Interactive Analysis: Agents can run code where data already exists and see the plots they generate.
• Manuscript Auditing: Includes "Reviewer Zero" for auditing claims, reconciling numeric values against code output, and verifying references via Crossref.
• Multi-Agent Coordination: A typed event log allows multiple agents to coordinate tasks, share state, and reach consensus on plans.
• Research Workflows: Built-in prompts for systematic review screening, grant panel simulations, and responding to peer reviewers.
• Session Management: Supports session checkpoints to snapshot and restore the global environment.
• IDE Integration: Works as an addin for RStudio and integrates with Positron.