# vndee/llm-sandbox

Lightweight and portable LLM sandbox runtime (code interpreter) Python library.

Repository: https://github.com/vndee/llm-sandbox
Canonical: https://ross.abutalabs.com/products/llm-sandbox
Homepage: https://vndee.github.io/llm-sandbox/
Language: Python
License: MIT
License Family: permissive
Topics: code-generation, code-interpreter, large-language-models, llm-sandbox
Last push: 2026-09-02T06:27:59+00:00

## Health v2 (maintenance only)
Score: 90/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 100, release rhythm 96, longevity 56
- inputs: {"age_days": 797, "days_push": 0, "days_rel": 30, "gap_med": 1.5, "n_releases_24m": 45}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1118, forks 104 (observed 2026-09-03T02:15:20.379685+00:00)

## What it is
LLM Sandbox is a lightweight Python library for safely executing LLM-generated code in isolated container environments with support for Docker, Kubernetes, and Podman backends. It offers multi-language code execution, security policies, resource limits, and an MCP server integration for AI assistants like Claude Desktop.

## Use cases
- run code generated by LLMs safely in an isolated sandbox
- execute AI-generated Python or JavaScript with automatic dependency installation
- add a secure code interpreter to an AI agent or LangChain pipeline
- let Claude Desktop execute code via an MCP server
- capture plots and artifacts produced by sandboxed code
- run untrusted code with CPU, memory, and network restrictions

## When to choose
- you need to execute untrusted LLM-generated code without risking the host system
- you want a Python library with pluggable container backends (Docker, Kubernetes, Podman)
- you're building AI agents that need a code interpreter tool
- you need multi-language execution with on-the-fly dependency management

## When to avoid
- you need a fully managed cloud code-execution service rather than self-managed containers
- your environment cannot run containers (no Docker/Podman/Kubernetes)
- you only need to run trusted code with no isolation requirements

## Facets
- artifact type: library
- maturity: active
- function: security, sdk, mcp, developer-tools
- domain: large-language-models, security, developer-tools
- platform: python, cross-platform
- tags: code-interpreter, sandbox, llm-generated-code, container-backends, mcp-server, ai-agents, docker, kubernetes

## Member repositories
- vndee/llm-sandbox (main) score 90

## Provenance
- Observed fields: from GitHub, fetched 2026-09-03T02:15:20.379685+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:42:53.717375+00:00, confidence not recorded.
  - readme: https://github.com/vndee/llm-sandbox (fetched 2026-09-03T02:15:20.379685+00:00, sha e47bce63788c)
  - homepage: https://vndee.github.io/llm-sandbox/ (fetched 2026-08-29T12:46:49.749597+00:00, sha 4f42cbd5aa07)
  - registry_pypi: https://pypi.org/pypi/llm-sandbox/json (fetched 2026-08-29T12:46:49.758678+00:00, sha 16638c5ae578)
- Data as of 2026-08-30T08:39:29.467469+00:00.