Yes. Agent Zero is open source under the MIT licence, so running it on your own server costs nothing beyond the server. What your model provider charges for its API is separate.

How do I sign in to Agent Zero?

With the username admin and the password the panel generated, which you find in the credentials card on the app's Container page. Agent Zero asks for them from the first start, because the agent runs shell commands and code on your behalf, and its developers warn never to expose it without a login. If you change the login or password in Agent Zero's Settings, the change is saved on the data volume and survives restarts, but the credentials card still shows the original password.

Which AI models can I use?

No chat model ships with the recipe. Agent Zero supports many providers, including OpenRouter, OpenAI, Anthropic, Google, DeepSeek, Mistral, Groq, xAI, Azure OpenAI and Amazon Bedrock, sign-in with a Codex/ChatGPT or GitHub Copilot account, and servers you run yourself such as Ollama, LM Studio or vLLM. You enter the provider and its API key in the set-up that opens with your first message, or later in Settings. Only the small default embedding model, which indexes memory and knowledge, runs on your server's CPU.

How much RAM does Agent Zero need?

Agent Zero's VPS deployment guide asks for at least 2 GB of RAM and 1 vCPU, and recommends 4 GB or more and 2 or more vCPUs. The recipe gives the container a 2 GB memory limit, which you can raise under Memory limit.

Can the agent run commands on my server?

Only inside its own container. That is where it runs shell commands and code, next to its own SearXNG, cron and an internal SSH server. Only the web UI is published, and the container has no access to Docker and no extra privileges. With the login on, as it is by default, the password protects the app; in case the login is ever turned off, the recipe also limits Agent Zero to browsers on your app's own HTTPS address.

Where is my data kept, and how do I update?

On your server, in one Docker volume mounted at /a0/usr: settings, chats, memory, knowledge, projects and uploads, so they survive container updates. Update Agent Zero by changing the image in the panel, not with the in-app Self Update, which changes files outside that volume and is lost when the container is re-created. Before big changes, create a backup in Settings under Backup & Restore.

Agent Zero on your own server, in one click.

Run Agent Zero, an open-source AI agent that works in its own Linux environment with code, a terminal, files, a browser and memory, on a server you own. You pick the model: OpenRouter, OpenAI, Anthropic, Google and many more. ServerAvatar runs it in Docker, puts a login in front of it from the first start and turns HTTPS on for you.

One container
Login from the first start
Your server, your agent

Deploy Agent Zero See how it installs

Section 01What is Agent Zero

What is Agent Zero?

Agent Zero is an open-source AI agent framework, published by agent0ai on GitHub. You give it a task in a web chat, and it does the work in its own Linux environment inside the container: it writes and runs code, uses the terminal, reads and writes files, and searches the web through its built-in SearXNG. It can hand parts of a job to subordinate agents, and it keeps a memory of useful facts between chats.

No chat model ships with the recipe: you connect a provider such as OpenRouter, OpenAI, Anthropic or Google, or a model server you run yourself, such as Ollama. Chats, memory, projects and files stay in a volume on your server. What you send in a chat, with the context the agent needs, goes to the provider you pick.

Free and open source (MIT)
No chat model bundled: connect your own provider
Shell and code stay inside its container

Section 02Features

What you can do with Agent Zero

Projects
Give a chat its own workspace with files, instructions, memory, secrets and model settings, then pick the project in the chat's project picker.
Memory you can curate
Agent Zero remembers useful facts and imported knowledge. In the Memory dashboard you search, review, edit and delete what it keeps.
Scheduled tasks
Under Tasks, create scheduled, planned or ad-hoc jobs, such as a weekly report, each with its own prompt and an optional project.
Subordinate agents
The main agent can create subordinate agents for focused parts of a larger job. Each keeps its own context and reports back when done.
Browser and Linux desktop
Watch the agent browse in the Browser surface and annotate pages, or open its own XFCE desktop for Linux GUI apps and LibreOffice files.
Skills, MCP and plugins
Pin skills to a chat, connect MCP servers, and install community plugins from the Plugin Hub in the Web UI.

Section 03Requirements

What you need

A fresh Ubuntu server you can log in to as root. The installer checks all of this before it changes anything.

Agent Zero on Docker
Requirement	Agent Zero on Docker
Operating system	Ubuntu 24.04 or 26.04 LTS (x86-64)
CPU	1 core or more
RAM	2 GB or more
Disk	25 GB or more
Containers	Agent Zero 2.14 (SearXNG built in)
Ports	80 and 443 open

CPU and RAM from Agent Zero's VPS deployment guide: at least 1 vCPU and 2 GB, with 2 or more vCPUs and 4 GB or more recommended. Disk: about 18.5 GB is in use once the panel and Agent Zero are installed on our test server; 25 GB is Ubuntu's recommendation.

Section 04Install

How to install Agent Zero

Every step happens in the ServerAvatar panel. No terminal needed.

Pick Agent Zero in 1-Click Recipes
On your server, open Applications, click Create application and search for Agent Zero under 1-Click Recipes.
Name it and choose a domain
Give the app a name and use your own domain, or a temporary one that works right away. The panel creates a system user for it.
Create it
Click Create application. The panel starts Agent Zero, turns HTTPS on and shows the admin password it generated in the credentials card on the Container page. Copy it somewhere safe.
Sign in and connect a model
Open the app's domain. Agent Zero's login page asks for a Username and a Password: enter admin and the password from the panel, and the dashboard opens. No model is bundled: the model set-up opens when you send your first message. Pick a provider such as OpenRouter, OpenAI or Anthropic, paste its API key, choose a main model and a utility model for quick internal tasks, and click Finish setup. You can change the models later in Settings under Agent Settings, Models, and the keys under External Services, API Keys.

Section 05FAQ

Questions,answered.

Can't find the answer you're looking for? Reach out to our customer support.

Contact support

Risk free

Deploy your first appin 10 minutes.

Connect a server, pick an application and go live, with SSL and backups already in place.

Start free Plans & pricing

No credit card required
Free 4-day trial
Deploy in the next 10 minutes

Agent Zero on your own server, in one click.

What is Agent Zero?

What you can do with Agent Zero

Projects

Memory you can curate

Scheduled tasks

Subordinate agents

Browser and Linux desktop

Skills, MCP and plugins

What you need

How to install Agent Zero

Pick Agent Zero in 1-Click Recipes

Name it and choose a domain

Create it

Sign in and connect a model