Start Agent Zero, open the Web UI, and give it a small, concrete job. If models need setup, Agent Zero keeps your first message while you configure them.
Use A0 Launcher for guided desktop setup. It can prepare the local runtime, install Agent Zero, and open local or remote Instances.
For a terminal installation, use A0 Install:
macOS / Linux:
curl -fsSL https://bash.agent-zero.ai | bashWindows (PowerShell):
irm https://ps.agent-zero.ai | iexFollow the installer prompts for port and authentication. Open the Web UI URL shown by Launcher or the terminal. See the Installation Guide for downloads, direct Docker, shared folders, and updates.
The welcome screen has a message box. Type a request there and press Enter; Agent Zero creates the chat for you. New Chat in the sidebar also starts a fresh conversation.
Try a small task with an obvious result:
Create a one-page Markdown plan for a 15-minute team kickoff.
Include a goal, three agenda items, and next steps.
Save it as /a0/usr/workdir/docs-demo-20261030/kickoff-plan.md.
If models are not configured, the chat offers Cloud provider, AI account, or Local model. Choose a path and set the Main and Utility models. If your provider is missing, use the provider-plugin link. The Codex guide covers ChatGPT account sign-in; the OpenCode Go guide covers installing a provider plugin.
Click Finish setup, then Continue to close the wizard. Agent Zero resumes the message it was holding in the same chat; you do not need to send it again. See First-Run Onboarding for the full walkthrough.
Tip
If the welcome message box is hidden, use Show all dismissed at the bottom of Welcome. See Customize Welcome.
Watch the response and expand tool rows when you want to see what Agent Zero did. The example below writes a Markdown plan and returns a link to the file.
Open the returned file link to check the plan. If no reply arrives, check that setup finished and read any provider error in the chat. If the agent replies but does not provide a file, ask it to save the plan and return the link.
Use Pause Agent and Resume Agent to pause and continue a conversation. Messages sent during a run enter its queue; use Send now on a queued item when you want to send that correction immediately. The Usage Guide explains the queue, Stop, Nudge, and other controls.
Agent Zero normally works inside its container. To give it access to host files, use Launcher Host access, the A0 CLI Connector, or an explicitly shared local folder.
Launcher maintains its supporting CLI. For terminal chat or a connection from another computer, run the A0 CLI installer on that computer:
macOS / Linux:
curl -LsSf https://cli.agent-zero.ai/install.sh | shWindows (PowerShell):
irm https://cli.agent-zero.ai/install.ps1 | iexThen run a0 and connect to your running Agent Zero instance. The
connector guide covers host selection and access
controls.
Now that you've run a simple task, you can experiment with more complex requests. Try asking Agent Zero to:
- Create a project for a focused workspace.
- Use the built-in Browser to research, screenshot, or annotate a page.
- Open the Desktop when you want Linux GUI apps or LibreOffice Cowork.
- Review Memory when Agent Zero seems to keep the wrong assumption.
- Connect A0 CLI when Agent Zero should work on host-machine files.
- Use + → Skills when you want to pin or remove a skill in the current chat.
- Choose an Agent Profile for a repeatable role, instructions, model preset, and capability permissions.
- Use the first model dropdown when you want to choose or edit Model Presets.
- Attach files and ask for a summary, edit, or conversion.
- Create a scheduled task for recurring work.
- Explore plugins when you need installed integrations or custom UI features.
Explains the built-in Browser, live Browser Canvas, screenshots, annotations, host-browser mode through A0 CLI, and Chrome extensions.
Shows the right-side Canvas Linux desktop, the New menu for Markdown/Writer/Spreadsheet/Presentation files, and LibreOffice Cowork.
Explains how to search, edit, delete, export, and curate memories before stale context starts steering the agent.
Shows the chat input + menu, the Skills selector, and how active skills are added to prompt protocol.
Shows how to switch profiles, create one in Easy mode, edit prompts in Advanced mode, and set Tool, MCP, Skill, and project-availability policies.
Explains presets as simple named shortcuts for model setups.
Provides more in-depth information on chat controls, tools, projects, tasks, and backup/restore.

