Skip to content

Odysseus, in plain language

What it is

Odysseus is a self-hosted "AI workspace" you run on your own computer or server. Instead of using ChatGPT or another company's website, you run Odysseus yourself, point it at whatever AI model you want (a local model on your own machine, or a paid API like OpenAI/Anthropic), and get a chat assistant plus a bunch of connected tools: email, calendar, notes, documents, research, an image gallery/editor, and more.

Think of it like running your own private version of a ChatGPT-style app, except it can also read your email, manage your calendar, edit documents, and run commands on your computer if you let it - because it's yours, running on hardware you control.

Why that matters

Because it runs on your own machine and can do powerful things (read files, send email, run shell commands), it is built more like an admin tool than a public web app. The project's own security docs describe it as "treat it like an admin console" - it's meant for you (and people you trust), not for strangers on the internet.

The main pieces, without the jargon

  • Chat - talk to an AI model, with memory and access to tools.
  • Agents/tools - the AI can be given permission to search the web, read and write files, run shell commands, send email, and more. Powerful features are restricted to admin accounts by default.
  • Research - point it at a topic and it reads multiple web pages and writes you a report.
  • Documents, Notes, Calendar, Email - everyday productivity features, all stored locally.
  • Cookbook - helps you figure out which local AI models will actually run well on your specific computer's hardware, and can download/run them for you.
  • Gallery/Editor - an image gallery with an AI-assisted image editor.

What "self-hosted" means for you

  • Your data stays on your machine (or wherever you deploy it) unless you connect it to an external service (like an email account or a paid AI API).
  • You are responsible for keeping it secure: don't expose it to the open internet without authentication, keep it behind something like a VPN or a trusted access proxy if you do expose it remotely, and keep your login credentials strong.
  • It talks to whatever AI provider you configure - local (private, free, needs decent hardware) or a paid API (easier, costs money per use, sends your data to that provider).

Where to go next