agent-zero: Give AI Agents a Linux Workbench

Overview

Agent Zero is an extensible framework for running AI agents in a Dockerized Linux environment. It addresses the limits of chat-only agents by giving them tools for working with files, desktop applications, web pages, and editable documents, while keeping their actions visible to the user.

It is aimed at developers and technically capable users who want to customize an agent's tools and workflows or apply it to practical work such as coding, research, and document production. The A0 CLI Connector can extend a running instance to trusted files and shell access on a host machine.

Key Features

• Linux desktop environment for using terminals, files, and GUI applications.
• Built-in browser with DOM inspection and annotations that can guide page changes or UI reviews.
• Live collaboration on Markdown and LibreOffice documents.
• Projects that separate workspaces, instructions, memory, secrets, and model settings.
• Extensible through plugins, custom tools and prompts, skills, MCP, and A2A connectors.
• Multi-agent delegation for splitting work across focused subagents.
• Optional A0 CLI Connector for connecting an instance to host-machine files and shell workflows.
• Workspace snapshots with diff inspection and rollback for Agent Zero-managed files.

Use Cases

• Developers can use it to inspect repositories, make changes, run tests, and work with local files through the host connector.
• Designers and frontend developers can annotate reference pages and turn observations into implementation or review instructions.
• Analysts can use the desktop and document tools to gather information and produce editable spreadsheets or reports.
• Teams managing separate client or project work can isolate files, instructions, memory, and credentials in individual projects.

Getting Started

The README recommends Docker. With Docker available, start an instance with:

docker run -p 80:80 -v a0_usr:/a0/usr agent0ai/agent-zero

Then open the Web UI and configure an LLM provider. See the README and installation guide for setup options and details.

Alternatives

• nanobot: Nanobot centers on a personal agent accessed through chat, browser, or terminal, while Agent Zero emphasizes a Dockerized Linux desktop and tools for working in it.
• deer-flow: DeerFlow focuses on coordinating agents for long-running research, coding, and content tasks, while Agent Zero provides agents with a general-purpose desktop environment.

Considerations

• Docker is required for the documented container setup. The project also describes using it on systems where Docker runs.
• The agent can interact with a real desktop and, through the CLI Connector, host files and shell commands. These capabilities require careful access controls and review, especially for sensitive data or production systems.
• The repository input reports its license as NOASSERTION, so verify licensing terms before adopting or redistributing it.
• Agent Zero is a framework with configurable providers, tools, and extensions. Expect to configure an LLM provider and choose which integrations and permissions to enable.