Claude REPL Server

You deploy an MCP server that exposes a Claude-based REPL with Python execution, shell commands, and safe file manipulation. It helps you run quick Python snippets, manage shell tasks, and perform text transformations in files from a consistent, scripted interface.

How to use

You connect your MCP client to the Claude REPL Server using the provided stdio interface. This server runs Python in a fresh or persistent state depending on the tool you choose. Use the one-off Python tool for isolated computations, or the session-based Python tool when you want to keep variables between executions. You can also run shell commands or perform safe file edits through the built-in tools.

How to install

Prerequisites you need before installing: a Python 3.x environment and a compatible tool for managing dependencies (Poetry or the standard Python package installer). You will also need access to the configuration step to connect this MCP server to your Claude client.

# Option 1: install dependencies with Poetry
poetry install

# Option 2: install dependencies with pip (from the pyproject.toml)
uv pip install -r pyproject.toml

{
  "mcpServers": {
    "repl": {
      "command": "/path/to/python/env/bin/python",
      "args": ["-m", "repl"]
    }
  }
}

Configuration details

Configure the Claude REPL Server in your MCP client to connect via the stdio channel using the Python interpreter and the REPL module. The config below shows the exact executable and arguments to start the REPL tool in a dedicated environment.

{
  "mcpServers": {
    "repl": {
      "command": "/path/to/python/env/bin/python",
      "args": ["-m", "repl"]
    }
  }
}

Notes and tips

• The Python tools support two modes: one-off Python for isolated runs and session-based Python for persistent state during a session. Sessions remain active for a short period after you stop sending commands, so you can continue working without reloading the environment.

• Shell tools execute commands and return immediately for quick tasks. For long-running commands, you receive a task ID and can monitor progress with the status tool.

• File tools enable safe, UTF-8 compliant text transformations while preserving file structure. Use them for multi-line, pattern-based edits.

Available tools

python

One-off Python tool that runs code in a fresh environment for quick calculations and tests; Pandas and PyArrow are supported.

python_session

Session-based Python tool that maintains persistent state for multiple executions within a session, useful for data analysis workflows.

shell

Command execution tool that runs shell commands; short tasks return immediately while long-running commands switch to async mode and provide a task ID for tracking.

shell_status

Monitor the progress of long-running shell commands; retrieve outputs, execution time, and results upon completion.

perl

File modification tool for safe text processing with multi-line transformations and UTF-8 support; preserves file structure.