Supports containerized deployment of the MCP server, with configuration options for connecting Modal resources.
Utilizes .env files for configuration management, storing API tokens and server settings.
Provides sandbox management in the cloud with GPU support, including launching isolated Python environments, managing packages, configuring resources, and executing remote commands.
Creates customizable Python environments with support for multiple Python versions and package management.
MCP4Modal Sandbox
A powerful Model Context Protocol (MCP) server that provides seamless cloud-based sandbox management using Modal.com. This project enables LLMs and AI assistants to spawn, manage, and interact with isolated compute environments in the cloud with full GPU support.
Features
Core Sandbox Management
Launch Sandboxes: Create isolated Python environments with custom configurations
Terminate Sandboxes: Clean resource management and controlled shutdown
List Sandboxes: Monitor and track active sandbox environments
App Namespacing: Organize sandboxes within Modal app namespaces
Advanced Configuration
Python Versions: Support for multiple Python versions (default: 3.12)
Package Management: Install pip and apt packages during sandbox creation
Resource Allocation: Configure CPU cores, memory, and execution timeouts
Working Directory: Set custom working directories for sandbox environments
GPU Support
Comprehensive GPU support for machine learning and compute-intensive workloads:
T4: Entry-level GPU, ideal for inference workloads
L4: Mid-range GPU for general ML tasks
A10G: High-performance GPU for training (up to 4 GPUs)
A100-40GB/80GB: High-end GPUs for large-scale training
L40S: Latest generation GPU for ML workloads
H100: Latest generation high-end GPU
H200: Latest generation flagship GPU
B200: Latest generation enterprise GPU
File Operations
Push Files: Upload files from local filesystem to sandboxes
Pull Files: Download files from sandboxes to local filesystem
Read File Content: View file contents directly without downloading
Write File Content: Create and edit files within sandboxes
Directory Management: Create, list, and remove directories
Command Execution
Remote Execution: Run arbitrary commands in sandbox environments
Output Capture: Capture stdout, stderr, and return codes
Timeout Control: Configure execution timeouts for long-running tasks
Performance Metrics: Track execution time and resource usage
Security & Environment Management
Secrets Management: Inject environment variables and secrets
Predefined Secrets: Reference existing secrets from Modal dashboard
Volume Mounting: Attach persistent storage volumes
Isolated Environments: Complete isolation between sandbox instances
Transport Options
stdio: Direct command-line interface (default)
streamable-http: HTTP-based communication
SSE: Server-Sent Events for real-time updates
rerequisites
Python 3.12+
Modal.com account and API key
Environment variables configured (see Configuration section)
Installation
Using UV (Recommended)
Using Docker
Build the Docker Image
Run with stdio Transport (Default)
Configuration
Environment Variables
Create a .env
file in the project root:
Modal.com Setup
Create an account at Modal.com
Generate API tokens from your Modal dashboard
Configure the tokens in your environment variables
Integration with Claude Desktop
Add to your Claude Desktop configuration:
uvx
docker
Available Tools
The MCP server provides 11 tools for comprehensive sandbox management:
launch_sandbox
- Create new Modal sandboxes with custom configuration (Python version, packages, GPU, secrets)terminate_sandbox
- Stop and clean up running sandboxeslist_sandboxes
- List all sandboxes in an app namespace with their statusexecute_command
- Run shell commands in sandboxes and capture outputpush_file_to_sandbox
- Upload files from local filesystem to sandboxespull_file_from_sandbox
- Download files from sandboxes to local filesystemlist_directory_contents
- List contents of directories within sandboxesmake_directory
- Create directories in sandboxesremove_path
- Remove files or directories from sandboxesread_file_content_from_sandbox
- Read file contents directly from sandboxeswrite_file_content_to_sandbox
- Write content to files within sandboxes
This server cannot be installed
A Model Context Protocol server that enables LLMs and AI assistants to create, manage, and interact with isolated cloud-based Python environments with GPU support on Modal.com.
Related MCP Servers
- AsecurityAlicenseAqualityA Model Context Protocol server that enables AI assistants like Claude to interact with Google Cloud Platform environments through natural language, allowing users to query and manage GCP resources during conversations.Last updated -9246166MIT License
- -securityFlicense-qualityA Model Context Protocol server that enables AI assistants like Claude to perform Python development tasks through file operations, code analysis, project management, and safe code execution.Last updated -5
- AsecurityFlicenseAqualityA Model Context Protocol server that allows LLMs to interact with Python environments, enabling code execution, file operations, package management, and development workflows.Last updated -9
- AsecurityAlicenseAqualityA secure Model Context Protocol server that allows AI assistants and LLM applications to safely execute Python and JavaScript code snippets in containerized environments.Last updated -2145MIT License