Install
$ agentstack add mcp-boxlite-ai-boxlite-mcp ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
boxlite-mcp
MCP server providing isolated sandbox environments for AI agents.
The Problem
Today's AI apps don't just generate text – they write and run code, call tools, read your files, and hit the network. Running all of this directly on your machine creates real risks:
- Security - Malicious or buggy code can damage your system
- Privacy - Sensitive files and credentials are exposed
- Reliability - Runaway processes can consume resources or crash your machine
BoxLite solves this by giving AI agents their own isolated VM – full freedom inside, complete safety outside.
Powered by BoxLite
BoxLite is an embeddable virtual machine runtime that follows the SQLite philosophy - simple, lightweight, and zero-configuration.
Why BoxLite?
- Hardware-level isolation - True VM security, not just containers. Your AI agent runs in a completely isolated environment.
- No daemon required - Unlike Docker, BoxLite doesn't need a background service. Just import and use.
- Embeddable - Designed to be embedded directly into your applications, like SQLite for compute.
- Fast startup - VMs boot in seconds, not minutes.
- Cross-platform - Works on macOS and Linux.
Use Cases
- AI Agent Sandboxing - Let AI agents execute code, browse the web, and use applications safely
- Secure Code Execution - Run untrusted code without risk to your host system
- Browser Automation - Headless browser with CDP for web scraping and testing
- Development Environments - Disposable, reproducible dev environments
Demo
https://github.com/user-attachments/assets/0685d428-64e4-4a68-adfe-c24dc0dc5ae8
Available Tools
| Tool | Description | |------|-------------| | computer | Full Ubuntu desktop with XFCE. Anthropic computer use API compatible. | | browser | Chromium browser with CDP endpoint for Puppeteer/Playwright/Selenium | | code_interpreter | Python code execution sandbox | | sandbox | Generic container for running shell commands |
Quick Start
Claude Code
claude mcp add boxlite -- uvx boxlite-mcp
Claude Desktop
Add to your Claude Desktop configuration (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
{
"mcpServers": {
"boxlite": {
"command": "uvx",
"args": ["boxlite-mcp"]
}
}
}
Manual Installation
pip install boxlite-mcp
Development
git clone https://github.com/boxlite-labs/boxlite-mcp.git
cd boxlite-mcp
uv sync --extra dev
uv run pytest
Connect Local Server to Claude Code
claude mcp add boxlite -s local \
-e BOXLITE_HOME=$HOME/.boxlite-mcp \
-- uv run --directory /path/to/boxlite-mcp python -m server
> Note: Environment variables must be passed via -e, not inline in the args. MCP stdio servers bypass the shell, so VAR=value command syntax won't work.
Connect Local Server to Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):
{
"mcpServers": {
"boxlite": {
"command": "uv",
"args": ["run", "--directory", "/path/to/boxlite-mcp", "python", "-m", "server"],
"env": {
"BOXLITE_HOME": "/path/to/.boxlite-mcp"
}
}
}
}
License
Apache-2.0
Source & license
This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: boxlite-ai
- Source: boxlite-ai/boxlite-mcp
- License: Apache-2.0
- Homepage: https://boxlite-labs.github.io/website/
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.