Verified project record

agent0ai/agent-zero

agent0ai/agent-zero is an open agent framework that gives an LLM a Dockerized Linux desktop, a browser with DOM annotation, live document cowork, and a bridge to the host machine. It targets work that requires operating real GUI software, browsing web pages, editing documents collaboratively, and executing actions in an isolated environment.

Project overview

Rather than limiting an agent to a chat window, the framework lets it drive a real XFCE desktop session, annotate live web pages as interactive directive surfaces, and co-edit Markdown documents with the user, addressing multi-step tasks that need genuine environment interaction.

Deployment: Refer to project documentation
License: License pending

Best for

Developers and users who need an agent to operate real GUI software, browse and annotate web pages, and co-edit documents inside a Dockerized Linux desktop.
Teams that want isolated workspaces, on-demand skills, agent profiles, model presets, and multi-agent delegation for research, coding, or analysis.

Key capabilities

Provides a real XFCE desktop session running in the container, allowing the agent to drive real GUI software, terminals, and files.
A built-in browser where users can click page elements to inspect, change, lift, or comment on them as actionable directives.
A rich Markdown editor built for simultaneous editing where agent and user edits are equal first-class operations on the same document.
Allows real-time creation, updating, saving, and verifying of ODT, ODS, and ODP files in LibreOffice Writer, Calc, and Impress.
A built-in hub to browse and install over 100 community plugins covering development, memory systems, tools, UI extensions, and workflows.
Enables agents to create subordinate agents to delegate research, coding, or analysis tasks and keep contexts focused.
Connects a running instance to the host machine via a terminal-native bridge, allowing it to work on real local files.
Allows isolating workspaces, loading skills on demand, changing the broader working style, and quickly switching model setups.

Limitations and risks

The framework is designed to run inside Docker or another isolated environment.
Snapshot history, diff inspection, travel, and revert are provided for /a0/usr workspaces, but Time Travel is not a replacement for Git or backups.
Granting A0 CLI Read+Write access and remote code execution poses security risks for untrusted machines; review this adoption check before enabling the host bridge.

Getting started

Setup is rated easy because the framework provides a guided desktop launcher, simple curl/PowerShell installation scripts, and a direct Docker run command. To start: run the Docker container or use the A0 Launcher, open the Web UI on port 80 or 5080, configure your LLM provider, and begin with a concrete task. No coding is required.

Evidence and sources

Open original GitHub project