Onlyboxes by Coooolfan is a self-hosted code execution sandbox platform for individuals and small teams. Built primarily with Go, it provides a control-plane console and an execution-plane worker architecture. This free mcp server tool connects to various AI systems, including LLM Chat Clients, through its exposed MCP tools and REST APIs. It enables secure, sandboxed environments for running code and managing interactive sessions, making it a valuable asset for AI-driven workflows on the mcp platform.

What does Onlyboxes connect to?

Onlyboxes connects to various AI systems, specifically LLM Chat Clients, through its exposed MCP tools and REST APIs. It supports HTTP as a transport protocol for its MCP endpoint, ensuring compatibility with any LLM Chat Client. The console component itself uses both gRPC and HTTP for its operations. This design allows Onlyboxes to act as a crucial bridge, enabling AI models to interact with external execution environments and data processing capabilities.

Core features and tools

Onlyboxes exposes specific callable tools for AI interaction. These include pythonExec for secure Python code execution within a sandboxed environment. It also offers terminalExec, providing stateful terminal sessions for interactive command-line operations. Additionally, readImage processes images, making them readable or interpretable by AI models. All these MCP tools are also accessible via a REST API, which supports an asynchronous task API for managing operations.

Who is it for?

Onlyboxes is designed for individuals and small teams requiring a self-hosted code execution sandbox platform. It suits those who need to connect AI systems, particularly LLM Chat Clients, to external execution and data processing capabilities. This tool is ideal for developers and researchers building AI applications that require secure, isolated environments for running code or interacting with system resources. Skip this if your project does not involve integrating LLM Chat Clients with external execution environments.

How to get started

Access to the /mcp and execution APIs requires configured tokens. The server authenticates using an access token created from the token management page, and workers authenticate to the console using a WORKER_ID and WORKER_SECRET over HTTP for the MCP endpoint.

Use cases

• Executing Python code in a sandboxed environment, allowing users or AI clients to run scripts securely and without impacting the host system. This ensures isolation for potentially untrusted code.
• Providing stateful terminal sessions for interactive command-line operations, enabling direct interaction with the execution environment for debugging or administrative tasks.
• Processing images to make them readable or interpretable by AI models, preparing visual data for advanced analysis and machine learning workflows.

Onlyboxes offers a specialized solution for integrating AI systems with secure code execution and interactive environments. Its architecture supports a range of AI-driven tasks, from running Python scripts to processing images for model interpretation. The platform's self-hosted nature provides control over the execution environment.