AGENTOS DOCS / GETTING STARTED

Introduction

AgentOS is an autonomous AI agent that runs on your own hardware. These docs cover installing it, connecting a model, and putting your agents to work.

What is AgentOS?

AgentOS runs a persistent background server on your computer. That server handles agent processing, memory, scheduling and tool execution. You work with it through a web interface in your browser at http://localhost:3456, or from your phone, your editor, the command line and other agents.

Each workspace is a folder on your machine with its own agent. Every agent has its own memory, model, job queue and settings. An agent can read and write files, search the web, run commands, operate your browser and desktop, call connected services, and run work on a schedule. You decide how much it does on its own.

How it fits together

THE SERVERA background process that keeps your agents running 24/7, even when the browser is closed.
THE MODELAny Ollama, llama.cpp or OpenAI-compatible model, local or remote. You choose it per workspace.
TOOLS & SKILLSMore than a hundred built-in tools, plus skills, plugins, MCP servers and connectors that you add.
PLAIN FILESNo database. Memory, jobs and config are readable files in .agentos/ folders.

Start here

1 · INSTALL & ACTIVATEDownload AgentOS, start the server, and enter your activation token. 2 · CONNECT A MODELPoint AgentOS at Ollama, llama.cpp or an OpenAI-compatible API. 3 · CREATE A WORKSPACEGive your agent a folder, a name and a purpose. 4 · START CHATTINGDescribe what you want done, and learn the slash commands.

Popular topics

// IN-APP HELP Type /help in any chat to see every command, tool, skill and plugin available to that agent. Type /help <topic> (for example /help /plan or /help web_search) for details and examples.