LLM Sandbox is a Python library and MCP server for running code generated by large language models inside isolated containers. It is used when AI-generated programs need controlled access to languages, dependencies, computing resources, or networks without directly reaching the host system. The catalogue add-ons help agents use the sandbox.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/vndee/llm-sandbox/claude-mdgit clone --depth 1 https://github.com/vndee/llm-sandboxWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/vndee/llm-sandbox/claude-md)<a href="https://agentmods.dev/instructions/vndee/llm-sandbox/claude-md"><img src="https://agentmods.dev/badge/instructions/vndee/llm-sandbox/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.02639 | $0.02639 |
| Opus 5 | $0.01319 | $0.01319 |
| Sonnet 5 | $0.00528 | $0.00528 |
| Haiku 4.5 | $0.00264 | $0.00264 |
Grade A, and why
llm-sandbox CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 351 lines — stays where its author put it; the contents beside it link to each section on GitHub.
LLM Sandbox - AI Assistant Guide
This document provides context and guidance for AI assistants working with the LLM Sandbox codebase.
Project Overview
LLM Sandbox is a lightweight and portable sandbox environment designed to run Large Language Model (LLM) generated code in a safe and isolated mode. It provides secure execution environments for AI-generated code while offering flexibility in container backends and comprehensive language support.
Version: 0.3.13 License: MIT Python Support: 3.10+ Documentation: https://vndee.github.io/llm-sandbox/ Repository: https://github.com/vndee/llm-sandbox
Core Architecture
Key Components
-
Session Management (
llm_sandbox/session.py)SandboxSession: Basic code execution in isolated containersArtifactSandboxSession: Execution with automatic artifact capture (plots, visualizations)InteractiveSandboxSession: Stateful sessions maintaining Python interpreter statecreate_session(): Factory function for session creation
-
Container Backends (
llm_sandbox/docker.py,llm_sandbox/kubernetes.py,llm_sandbox/podman.py)- Docker (default, most common)
- Kubernetes (enterprise-grade orchestration)
- Podman (rootless containers)
-
Language Handlers (
llm_sandbox/language_handlers/)- Python (
python_handler.py) - JavaScript/Node.js (
javascript_handler.py) - Java (
java_handler.py) - C++ (
cpp_handler.py) - Go (
go_handler.py) - R (
r_handler.py) - Ruby (
ruby_handler.py)
- Python (
-
Container Pooling (
llm_sandbox/pool/)- Pre-warmed container management for performance
- Thread-safe concurrent execution
- Configurable pool sizes and lifecycle management
-
Security (
llm_sandbox/security.py)- Security policies and pattern matching
- Code scanning for vulnerabilities
- Resource limits and network isolation
-
MCP Server (
llm_sandbox/mcp_server/)- Model Context Protocol integration
- Enables Claude Desktop and other MCP clients to execute code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago First seen · 351 lines · 2,639 tokens per session scan A a33cbf89080f
llm-sandbox CLAUDE.md is an instructions file published in the GitHub repository vndee/llm-sandbox (1,119 stars, last pushed 3d ago), licensed MIT. It adds 2,639 tokens to every session, about $0.0132 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
appwrite AGENTS.md
AGENTS.md instructions for appwrite/appwrite, covering appwrite, commands, stack, layout and libraries.
docker-zoneminder CLAUDE.md
Instructions for jantman/docker-zoneminder, covering claude.md, project overview, build and test, ci/cd and architecture.
MoleCode AGENTS.md
Instructions for AtomFlow-AI/MoleCode, covering agents.md — molecode, what this repo is, use the molecode skill, molecode-first rule and cli quick start.
MoaV AGENTS.md
AGENTS.md instructions for MotherofallVPNs/MoaV, covering what moav is, install & run, everyday operations, dashboards & access — what to hand the owner and where things live.
observability-mcp CLAUDE.md
Claude Code instructions for ThoTischner/observability-mcp, covering observability-mcp development guide, docker-first development, quick start, rebuild a single service and view logs.
wordpress-nginx-docker-compose CLAUDE.md
Claude Code instructions for urre/wordpress-nginx-docker-compose, covering claude.md, what this is, architecture, two .env files — don't confuse them and common commands.