SkillFoxx

MCP Python Interpreter

MCP Python Interpreter runs Python code in a chosen environment and reads or writes files inside the agent's working directory.

  • 104 ★ on the repository
  • Last commit
MCP server

Broad access

Why this level

  • The server runs arbitrary Python with the user's permissions and no isolation.
  • It can install packages, read and overwrite files and reach the network.
All reasons and checks

yzfly/mcp-python-interpreterAuthor:yzfly

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add mcp/mcp-python-interpreter

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Broad accessThe tool can change data, run commands or handle secrets.

  • The server runs arbitrary Python with the user's permissions and no isolation.
  • It can install packages, read and overwrite files and reach the network.

This is third-party code. Review the repository files before installing.

What it does

MCP Python Interpreter is an MCP server that lets an agent run Python: code as a string or a file, in the system environment or conda. Separate tools list installed packages and install new ones, and read or write text and binary files up to 1 MB. File operations and `run_python_code(save_as=...)` are limited to the folder from the `--dir` parameter unless `MCP_ALLOW_SYSTEM_ACCESS=1` is set. Those checks cover file handling only: the executed code itself runs with the server user's permissions.

Who it is for. Developers and analysts who want Claude Desktop or another client to test code and crunch data on their own machine.

Good fit when

  • An agent writes a CSV analysis script and has to run it right away.
  • Several conda environments exist and the right one must be chosen and a package installed.
  • `data_analysis.py` must run with arguments and its result be shown.
  • A client without its own shell must read and edit project files.

Not a fit when

  • Code comes from an untrusted source and there is no container or separate user.
  • A network block or timeout is required, but inline execution has no enforced timeout.
  • The agent's own shell, as in Claude Code, is enough.

Example request

Run `data_analysis.py` in the conda-base environment with --input=data.csv --output=results.csv and show the outcome.

MCP

Transport
stdio, http
Authentication
not required

Environment variables

MCP_ALLOW_SYSTEM_ACCESS
1 lifts the working-directory limit on file tools, 0 by default

Limitations

The server is not a sandbox. Inline and subprocess execution run arbitrary Python with the user's permissions and can touch files or the network regardless of the file checks. Inline execution has no enforced timeout. Run only trusted code or isolate the server in a container or a restricted account. Last commit: 1 November 2026. The `install_package` tool changes the chosen environment, so use a dedicated one for experiments. Compatibility is declared with MCP Python SDK 1.x and 2.x.

How to disable. Remove the `mcp-python-interpreter` entry from the client config and run `pip uninstall mcp-python-interpreter`. Delete files the agent created in the working folder and packages it installed by hand.

FAQ

Is it safe to run untrusted code?

No. The server does not isolate execution, so code gets the user's permissions. Untrusted code needs a container or a restricted account.

What does MCP_ALLOW_SYSTEM_ACCESS do?

At 0 the file tools work only inside `--dir`. A value of 1 lifts that restriction.

Which environments does it see?

System and conda ones. `list_python_environments` lists them and `--python-path` selects the interpreter.

Are there file limits?

Reading and writing text and binary files up to 1 MB; existing files are not overwritten without `overwrite=true`.

How do I run the server in a container?

The repository has a Dockerfile. A container gives the isolation the server itself lacks.

Discussion

No questions or reviews yet.

MCP servers

An MCP server that lets an agent read files, search code and run commands in a workspace

MCP serverBroad access1.2K

add-mcp writes an MCP server into the configs of 25 agents with one command: Claude Code, Codex, Cursor, VS Code, Zed and more

CLIExtended access321

Command-line inspector for HTTP MCP servers: lists tools, schemas, prompts and resources and calls any tool by hand.

CLIExtended access134

An MCP server that gives Claude Desktop and other clients file tools, a shell and sub-agents in the style of Claude Code

MCP serverBroad access305

A debugger MCP server: the agent sets breakpoints, steps through code and reads variables of a live program over DAP.

MCP serverBroad access107

A CLI compiler that reads app code, API specs and database schemas and generates a ready MCP server with tool permission checks.

CLIExtended access107

Comparison with similar tools

Cards that share the main search query, and how they differ.

ToolOfficialAuthenticationRepository starsMain difference
MCP Python Interpreter (this card)nonot required104
MCP Code Execution Servernonot required343runs code through MCP; MCP Python Interpreter selects conda environments, installs packages and reads files
Open Interpreterno68,879a standalone terminal agent; MCP Python Interpreter plugs into your client as a server

README in short

The README lists tools, `python://...` resources and three prompt templates, and its security section names the limits of the file checks plainly.

README badge

Report a problem