open-terminal: a self-hosted API-driven terminal for AI agents to execute code and manage files in sandboxed or bare-metal environments

What it solves

AI agents and automation tools often lack a secure, dedicated environment to execute code, manage files, and run shell commands. Open Terminal provides a self-hosted, API-driven terminal that acts as a remote shell for these agents, allowing them to interact with a filesystem and operating system without requiring direct access to the host machine.

How it works

Open Terminal exposes a REST API that allows external tools or AI assistants to send commands and manage files. It can be deployed in two primary modes:

• Docker (Sandboxed): Runs in an isolated container with pre-installed toolkits (Python, Node.js, git, etc.). This is the recommended mode for providing a safe playground for AI agents.
• Bare Metal: Installed via pip and runs directly on the host machine, providing the agent with full access to the host's actual files and environment.

It also supports a multi-user mode for small, trusted groups where each user is assigned a dedicated Linux account and home directory using standard Unix permissions.

Who it’s for

• AI Agent Developers: Those building agents that need to execute code or perform system-level tasks.
• Self-hosters: Users of Open WebUI who want to integrate a terminal environment for their AI assistants.
• Automation Engineers: People creating automation tools that require a remote, API-accessible shell.

Highlights

• Flexible Deployment: Supports both sandboxed Docker containers and bare-metal Python installations.
• Pre-bundled Tooling: The latest Docker image includes a wide array of data science libraries, ffmpeg, LaTeX, and the Docker CLI.
• ** uma own custom environments:** Ability to add apt, pip, and npm packages at startup via environment variables.
• Open WebUI Integration: Deep integration with Open WebUI, providing a built-in file navigation sidebar and support for both direct and system-level connections.
• API-First Design: Full interactive API documentation available at /docs upon deployment.