jupyagent 0.3.1

One-command installer for Jupyter + LLM agent sandbox

🤖 JupyAgent

JupyAgent is a CLI tool that installs a containerized environment running web UIs from Jupyter server, LLM agents (OpenCode and Claude Code), and a web terminal (ttyd).

Features

• 🛡️ Secure Sandbox: Agents operate in an isolated Docker container with controlled read/write access.
• ⚡ Real-time agentic coding in Jupyter: Watch agents write and execute code in Jupyter Lab in real-time via the MCP protocol.
• 🖥️ Integrated Stack:
  • Jupyter Lab: For code execution and notebook management.
  • OpenCode & Claude Code: AI coding agents with Jupyter MCP integration.
  • ttyd: Web-based terminal with no authentication required.
  • Dev Tools: git, vim, nano, build-essential, jq, htop, tree, and more.
• 🚀 One-command management: jupyagent, a TUI dashboard to manage the tool.
• 🔓 Seamless Access: No password prompts - click and start working.

Prerequisites

• Docker: Must be installed and running.
• uv: Recommended for running the tool.

Quick Start

Run JupyAgent directly:

uvx jupyagent

Usage

1. Setup: On first run, configure your workspace:

   • Context Path (Read-Only): Directory the agent can read (e.g., your data drive).
   • Workspace Path (Read-Write): Directory where the agent creates files (e.g., your project root).

2. Dashboard:

   • ▶️ Start/Stop Services: Toggle the background Docker container.
   • 📓 Open Jupyter Lab: Access the notebook interface (auto-authenticated).
   • 🤖 Open Opencode: Access the OpenCode Web UI to give instructions.
   • 💻 Open Web Terminal: Access the web terminal (instant access, no login).

Architecture

JupyAgent builds a custom Docker image combining:

• jupyter/base-notebook
• mcp-server-jupyter (with collaboration support)
• opencode CLI/Web
• claude CLI (Claude Code)
• ttyd web terminal
• Essential dev tools (git, vim, nano, build-essential, jq, htop, tree)