Agent Zero, the open source AI agent that writes and runs code, browses the web and uses tools inside its own Linux workspace, behind a private HTTPS login
This product is available in the build below. Open it for the operating system, cloud and version detail, or read its deployment guide.
| Product | Cloud | Operating system | Version | |
|---|---|---|---|---|
| Agent Zero | Azure | — | Standard | View · Guide |
Agent Zero is an open source, general purpose AI agent framework. You give it a task in a chat, and it plans, writes and runs code in a terminal, reads and edits files, searches and browses the web, keeps a memory of what worked and can hand sub tasks to subordinate agents. Everything the agent does happens inside its own Linux container, which acts as its workspace, and you watch each step live in the web interface. Agent Zero works with hosted models through your own API keys or with local models, exposes an external API and optional MCP and A2A endpoints for other tools, and is customised through prompts, skills, agent profiles and plugins.
It suits developers, analysts and operations teams who want a capable assistant that can actually carry out multi step technical work, in an environment they control.
Upstream Agent Zero runs with no login unless you configure one. cloudimg generates a login, a runtime identity and the API key derived from them for each virtual machine at first boot, keeps the password out of every log and command line, and starts nothing until that is done. nginx serves the interface over HTTPS with a certificate made for each VM, the container publishes only its web port and only to the VM itself, and a firewall stops the agent and its model server from reaching the Azure instance metadata service and WireServer. The official Agent Zero image is pinned by digest and pre pulled, an open weight model runs locally on the CPU so the agent works before you add any key, and every deployment comes with a step by step guide and 24/7 support.