Skip to content

📋 What you'll need

Before we dive into the magic of stateful kernels, let's make sure your environment is ready.


🛠️ Core Essentials

These are required to run KernelBox natively.

Tool Purpose
Python 3.10+ The heart of the system.
uv Our preferred (and lightning-fast) package manager.
ipykernel What actually runs the code your agent writes.

🌐 For the REST API

If you plan to use KernelBox over HTTP (e.g., via a FastAPI service):

Tool Purpose
FastAPI Powers our high-performance REST endpoints.
Uvicorn The engine that serves the API.

🐳 For Maximum Security

Highly recommended if you're running untrusted agent code.

Tool Purpose
Docker Keeps the agent code sandboxed and away from your host files.
Docker Compose Orchestrates the service with a single command.

💻 Supported Platforms

KernelBox is built to be cross-platform and has been tested on:

  • Windows (including specific fixes for Jupyter connection files)
  • macOS
  • Linux

Don't sweat the small stuff

If you're contributing to the project, our scripts/bootstrap.py will handle most of these for you!