Modal MCP Toolbox

Modal MCP Toolbox is an MCP server that provides AI assistants with access to cloud infrastructure tools hosted on the Modal platform. Designed for software developers and AI practitioners using clients like Claude Desktop and Goose, this server offloads resource-intensive compute tasks directly to serverless cloud environments. By integrating with an authenticated Modal account, the server enables connected language models to execute Python scripts inside secure, isolated sandboxes and render images through the FLUX model without requiring local GPU hardware. Users can dynamically generate, run, and evaluate arbitrary Python code in real time while capturing the stdout, stderr, and returned objects. Additionally, the tool provides direct prompt-to-image workflows by triggering cloud-based image diffusion jobs on Modal hardware. These capabilities eliminate the need to run untrusted code on a local workstation or maintain dedicated inference machines, transforming an MCP client into an interface for remote compute and image synthesis tasks.

Category: AI & LLM Tooling

Tags: cloud, compute, modal, serverless

Visit Modal MCP Toolbox

How to install and configure Modal MCP Toolbox

To set up Modal MCP Toolbox with Claude Desktop, ensure you have a Modal account, the configured Modal CLI, and uv installed. 1. Open your Claude Desktop configuration by navigating to Settings > Developer and clicking Edit Config. 2. Add the Modal MCP Toolbox configuration to the mcpServers object in claude_desktop_config.json: json { "mcpServers": { "modal-toolbox": { "command": "uvx", "args": ["modal-mcp-toolbox"] } } } 3. Save the file and restart Claude Desktop. For Goose, navigate to Settings > Add extension, set the command to uvx modal-mcp-toolbox, and save.

What you can do with Modal MCP Toolbox

• Running Python scripts inside remote cloud sandboxes to safely analyze untrusted code without executing it on your local machine. - Generating high-resolution images via the FLUX model using remote Modal compute directly from conversational client prompts. - Performing serverless data transformations and mathematical calculations using cloud compute resources from an AI assistant. - Automating complex Python code execution and debugging workflows directly through Claude Desktop or Goose chat interfaces.

Key facts

• https://github.com/philipp-eisen/modal-mcp-toolbox
• AI & LLM Tooling, Cloud & Infrastructure
• cloud, compute, modal, serverless

Part of MCP Servers

What tools are included in Modal MCP Toolbox?

The server includes two primary tools: run_python_code_in_sandbox, which lets your language model run Python code inside an isolated cloud sandbox, and generate_flux_image, which triggers image generation using the FLUX model running on Modal cloud infrastructure.

What prerequisites are needed to run Modal MCP Toolbox?

You need a registered Modal account, an active and configured Modal CLI login on your machine, the uv package manager installed, and an MCP-compatible client such as Claude Desktop or Goose.

Which MCP clients work with Modal MCP Toolbox?

Modal MCP Toolbox supports standard MCP clients. The repository provides explicit configuration instructions and verified testing for both the Claude Desktop application and the Goose agent extension system.

How do I install Modal MCP Toolbox?

You can run it on-demand using uvx. In your client configuration file, add modal-toolbox under the mcpServers key with the command set to uvx and the args set to modal-mcp-toolbox.