Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

Odysseus

PewDiePie's self-hosted AI workspace: chat, agents, deep research, documents, email and calendar behind one Docker Compose file.

Reviewed by ReviewRamble editors · Updated

View on GitHub Open source

Share this page:XLinkedInFacebookRedditWhatsAppEmail

Our rating

3.7/5

Odysseus packs more of a personal AI desk into one install than Open WebUI does, but it is under six months old and still moves faster than its docs.

What is Odysseus?

Odysseus is a self-hosted web workspace released by Felix Kjellberg (PewDiePie) on 31 May 2026. It puts chat and agents with tools, MCP, shell and memory beside a deep research mode, blind side-by-side model comparison, a document editor with AI edits, an IMAP/SMTP inbox with triage and reply drafts, notes, tasks and a CalDAV calendar. A Cookbook page recommends, downloads and serves local models that fit your hardware; cloud providers work through API keys.

The Docker Compose file starts four containers: the app, ChromaDB for vector memory, SearXNG for private web search and ntfy for reminders. Everything binds to 127.0.0.1 by default, and user data sits in a local data/ folder with a SQLite database.

The repo moved from pewdiepie-archdaemon to the odysseus-dev organisation and has about 95,710 GitHub stars and 1,998 forks. It changed from MIT to AGPL-3.0-or-later on 9 June 2026. There are no tagged GitHub releases; the last commit landed on 15 November 2026.

Is Odysseus worth it?

Most self-hosted chat front ends stop at chat. Odysseus also reads your mail, drafts replies, runs scheduled agent tasks and keeps your calendar, which makes it closer to a personal operating layer than to Open WebUI or LibreChat. Much of it reuses other projects: SearXNG for search, ChromaDB for memory, and Playwright MCP for browsing.

That breadth is also the security story. The setup guide calls it an admin console with shell access, file uploads, model downloads and stored email and API credentials. The defaults are sensible, with login on, the UI bound to localhost and shell tools limited to admins, but anyone exposing it beyond a laptop needs HTTPS, a reverse proxy and a check of data/auth.json for open signup.

The catch is maturity. There are 1,594 open issues and pull requests, no tagged releases, and the default branch is dev, which the docs call possibly unstable. The docs already drift: the Compose file publishes port 7011 while the setup guide still says 7000. The :latest Docker tag is re-pushed on every merge, so production users must pin an immutable tag. Outlook and Microsoft 365 mail do not work yet because only password IMAP is supported.

Our take: the scope is the reason to try it and the reason to wait. A single person on a home server with a GPU will get a lot from it; a team that needs stable upgrades is better served by Open WebUI for now.

Where it earns its place

  • A freelancer with a 24 GB GPU at home

    Keep email triage and research off cloud AI services

    Run the Compose stack, pick a local model in the Cookbook and connect the inbox over IMAP. Summaries and reply drafts stay on the machine; no fee beyond power, but local models will not match a frontier API.

  • A developer comparing models before choosing one for a project

    Test several models on the same prompts without bias

    Add local and API models, then use Compare for blind side-by-side answers. Free apart from the API calls made during the test.

  • A researcher who writes weekly literature notes

    Turn a question into a sourced report

    Deep Research runs multi-step searches through SearXNG, reads the sources and writes a report into the document editor. A fresh install ships with no model, so connect a local or API model first; quality follows the model you pick.

Before you commit

Setup
git clone, cp .env.example .env, docker compose up -d --build; the first admin password appears in docker compose logs odysseus. Native Linux, macOS and Windows installs are documented; on Apple Silicon, Docker cannot reach the GPU, so run natively.
Integrations
Ollama, vLLM, llama.cpp and OpenAI-compatible servers, cloud APIs, MCP servers, IMAP/SMTP mail, CalDAV, ntfy, SearXNG; API tokens and webhooks for automation.
Data and privacy
All data stays in data/ (SQLite, ChromaDB, uploads) on your host. It holds mail passwords and provider keys, and admins get shell access, so treat it as an admin console. Prompts leave the machine only if you add a cloud provider. AGPL-3.0-or-later.
Cost at scale
No licence cost. You pay for hardware or API calls; the app itself needs little, but local models want a GPU.
Limits
Chat uploads default to 10 MB; requirements are unpinned; local Kokoro voice needs Python 3.11 or 3.12 while the Docker image ships 3.14; Microsoft mail needs OAuth, which is not supported.

What’s good

  • Chat, research, email, notes and calendar in one self-hosted app
  • Local models through the Cookbook or any API provider
  • Localhost binding and login on by default
  • Private web search through bundled SearXNG

What’s not

  • No tagged releases; default branch is the unstable dev
  • 1,594 open issues and pull requests
  • Docs and Compose disagree on the port
  • No OAuth mail, so Outlook and Microsoft 365 fail

Best for

Individuals and homelab owners who want one private web app for AI chat, research and personal admin.

Odysseus alternatives

Odysseus: common questions

Is Odysseus worth it?
Odysseus packs more of a personal AI desk into one install than Open WebUI does, but it is under six months old and still moves faster than its docs. We rate it 3.7 out of 5.
How much does Odysseus cost?
Free and open source (AGPL-3.0-or-later). No paid tier; cloud model APIs bill you separately.
What are the downsides of Odysseus?
No tagged releases; default branch is the unstable dev; 1,594 open issues and pull requests; docs and Compose disagree on the port; no OAuth mail, so Outlook and Microsoft 365 fail.
Who is Odysseus best for?
Individuals and homelab owners who want one private web app for AI chat, research and personal admin.
How much does Odysseus cost for a team or in production?
No licence cost. You pay for hardware or API calls; the app itself needs little, but local models want a GPU.
Is Odysseus safe for company data?
All data stays in data/ (SQLite, ChromaDB, uploads) on your host. It holds mail passwords and provider keys, and admins get shell access, so treat it as an admin console. Prompts leave the machine only if you add a cloud provider. AGPL-3.0-or-later.
What are the best Odysseus alternatives?
The closest alternatives we have reviewed are Open WebUI, LibreChat and AnythingLLM.

Answers checked Nov 16, 2026.