Odysseus can do far more than chat. These tips show how to unlock the stronger parts: agents with tools and MCP, deep research with web search, email and calendar, handy extra features, and solid hardening. This assumes an installed Odysseus instance.
10 minIntermediateTested on Ubuntu 24.04, DockerUpdated 2026-06-18
In short
Enable the browser MCP, use web search and deep research, set up email and calendar, add optional packages, and review the security settings.
Odysseus registers a few built-in MCP servers at startup. The browser MCP for page navigation, screenshots, and vision only starts when its package is present locally. Enable it once:
Then restart Odysseus and the server registers at startup. Your agents can now open and read web pages, on top of shell, files, skills, and memory.
Web search runs through the bundled SearXNG, so requests do not go to an external service. With that, deep research works in multiple steps, reads sources, and writes a report. If you want DuckDuckGo as an additional source, add the optional ddgs package (see below).
Odysseus connects mailboxes over IMAP and SMTP, with triage, summaries, and reply drafts. Notes, tasks, and calendar can sync over CalDAV.
Outlook and Microsoft 365
Email accounts currently use username and password over IMAP/SMTP. Outlook and Microsoft 365 generally require OAuth, so a normal mailbox password will fail there.
The workspace itself is lightweight, the load comes from local models. If performance is not enough, either more CPU and RAM helps, or moving to a GPU server for fast local inference. How to set up GPU models is covered in Odysseus with a GPU. Upload limits like ODYSSEUS_CHAT_UPLOAD_MAX_BYTES you adjust in .env as needed.
More speed for local models: dedicated GPU servers from €99/month, from German data centers.
