# What Is Odysseus? Features, Use Cases and How to Host It (2026) > What is Odysseus? Learn how this self-hosted AI workspace runs chat, agents, memory and research, how it compares with Open WebUI and how to host it. Your team pastes contracts, code and customer emails into a hosted chatbot every day, and nobody is sure where those prompts end up. That worry is why so many people now ask: **what is Odysseus**? Odysseus is a free, open-source **self-hosted AI workspace** that puts chat, agents, memory, documents and research on a server you control. It does not ship a model of its own. Instead, it connects to the model you choose, such as a local **Ollama** server or an OpenAI-compatible API. If you already run models yourself, our guide to [**self-hosting Ollama**](/self-host-ollama/) pairs well with this one. In this guide, you will learn what Odysseus is, how it works and which features matter most. You will also see who it suits, how it compares with Open WebUI, LibreChat and AnythingLLM and how to host it without building the stack by hand. ## TL;DR: Odysseus in 60 Seconds Short on time? Here is the whole guide in a few lines: - **Odysseus** is an open-source, self-hosted AI workspace for chat, agents, deep research, documents, email, notes and calendar. - **It is young.** The first commit landed on May 31, 2026, and the project still changes almost every day. - **Bring your own model.** It connects to Ollama, vLLM, OpenAI or any OpenAI-compatible endpoint. No model is bundled. - **Your data stays on your server**: accounts, chats, uploads, memory and document indexes. - **It is free** under the AGPL-3.0-or-later license. You pay for the server and any API model you use. - **Compared with Open WebUI, LibreChat and AnythingLLM**, it packs more tools into one app, while they have a longer track record. - **On xCloud**, a one-click install runs Odysseus 1.0.3 with a private vector store, a generated admin password and HTTPS. ## What Is Odysseus? **Odysseus is a self-hosted web app that gives you a ChatGPT-style workspace, plus agents, research and everyday productivity tools, on hardware you control.** Its [**official GitHub repository**](https://github.com/odysseus-dev/odysseus) lists chat, agents, research, documents, email, notes, calendar and local model workflows as its scope. Think of it like a private office building for AI. Odysseus is the building (desks, filing cabinets, mailroom and meeting rooms), and the language models are the experts you invite in from down the hall or across town. Either way, you hold the keys and the filing cabinets never leave the building. A few things worth knowing: - The project first appeared on GitHub on **May 31, 2026**, from the pewdiepie-archdaemon account. Early coverage credits the YouTuber PewDiePie as its creator. It now lives in the **odysseus-dev** GitHub organization. - It launched under the MIT license and switched to **AGPL-3.0-or-later** on June 9, 2026. If you modify it and let others use it over a network, the license asks you to share your changes. - Interest has been strong. The repository showed about **94,900 GitHub stars as of November 1, 2026**. - The maintainers are open that it moves fast. Their roadmap puts bug fixing and fresh-install testing at the top of the list. - The `dev` branch gets the newest changes first. The `main` branch is the more curated one for steady use. | Detail | Odysseus | | --- | --- | | **What it is** | Self-hosted AI workspace (web app) | | **License** | AGPL-3.0-or-later | | **First public commit** | May 31, 2026 | | **Maintainer** | The odysseus-dev organization on GitHub | | **Backend** | Python (FastAPI) | | **Default database** | SQLite | | **Vector store** | ChromaDB, run as its own service | | **Models** | None bundled; connect Ollama, vLLM, OpenAI or another OpenAI-compatible endpoint | | **Install options** | Docker Compose (recommended) or native install on Linux, macOS and Windows | | **Minimum on xCloud's template** | 1 CPU core, 1.5 GB RAM, 10 GB disk | ## How Odysseus Works The recommended setup uses **Docker Compose**, which starts a few containers that work together. If Compose is new to you, our explainer on [**what Docker Compose is**](/what-is-docker-compose/) covers the basics in a few minutes. The default stack has three main parts: - **The Odysseus app:** a Python web app that serves the interface, stores accounts and chats in a SQLite database and calls your model. - **ChromaDB:** a vector database that stores your documents and memories as embeddings. As a result, search can match meaning rather than exact words. - **SearXNG:** a private metasearch engine that the upstream setup uses for web search in research and agent tasks. A small local embedding model turns your text into those vectors. It downloads from Hugging Face the first time you need it, then stays cached on disk. Here is the part that trips people up. **The app itself is light.** The project's [**setup guide**](https://github.com/odysseus-dev/odysseus/blob/main/website/setup.md) says local model serving is the heavy part, and that small hosts can connect to an API or a remote model server instead. ## Key Features of Odysseus These come from the project's README. Treat each one as evolving rather than finished. ### 1. Chat and Agents You get multi-turn chat with local or API models, plus agents that can use tools, MCP servers, files, a shell, skills and memory. Shell access is exactly what makes agents useful, and risky if deployed carelessly. ### 2. Cookbook: Models That Fit Your Hardware The **Cookbook** checks your hardware, recommends models that fit and can download and serve them. The roadmap names Cookbook reliability across different machines as a top priority, so expect rough edges. ### 3. Deep Research Deep Research runs multi-step web research, reads the sources and writes a report. It saves you from opening twenty tabs for one question. ### 4. Compare: Blind Model Testing Compare puts two models side by side without telling you which is which, then combines their answers. It is a quick way to judge whether a small local model is good enough. ### 5. Documents and Long-Term Memory The Documents editor is writing-first, with AI edits and suggestions for Markdown, HTML and CSV. Uploaded files and memories are indexed in ChromaDB, so the assistant can recall them later. ### 6. Email, Notes, Tasks and Calendar This is where Odysseus stands apart. It includes an IMAP/SMTP inbox with triage, tags, summaries, reminders and reply drafts. It also covers notes, todos, scheduled agent tasks and CalDAV calendar sync. ### 7. Extras: Gallery, Themes and 2FA Smaller tools round it out: an image gallery and editor, themes, presets, saved sessions, web search and two-factor sign-in. ## Who Is Odysseus For? ### Privacy-Minded People Who Want One AI Home If your notes, inbox and research live in five different apps, Odysseus pulls them into one place. Your accounts, uploads and memory stay on your own server. ### Local Model Tinkerers The Cookbook and Compare features suit anyone testing open models at home. You can see what fits your GPU, serve it and judge it blind against an API model. ### Small Teams That Share One Workspace A small team can share one agent workspace behind a single admin login. If each colleague needs a separate account with set permissions, compare user management first. Open WebUI and AnythingLLM both list multi-user access as a core feature. ### Teams Leaving a Hosted Chat Subscription Odysseus can replace the screen you chat in. However, if you connect a cloud API, your prompts still reach that provider. Full privacy only comes with a model you host yourself. ## Odysseus vs Open WebUI vs LibreChat vs AnythingLLM All four are good, open tools for running a private AI interface. The truth is less dramatic than most comparisons suggest: they share a lot, and the right pick depends on what you need beyond chat. | | **Odysseus** | **Open WebUI** | **LibreChat** | **AnythingLLM** | | --- | --- | --- | --- | --- | | **Main focus** | All-in-one workspace: chat, agents, research, email, calendar | Chat interface for local and cloud models | Chat for many AI providers in one place | Chat with your documents, plus agents | | **License** | AGPL-3.0-or-later | Open WebUI License (BSD-style, plus a branding condition) | MIT | MIT | | **Model connections** | Ollama, vLLM, OpenAI-compatible | Ollama, OpenAI-compatible | Anthropic, OpenAI, Google, Azure, custom endpoints | Local or cloud LLMs | | **Standout extras** | Cookbook, Deep Research, blind Compare, email, CalDAV | Roles and user groups, voice calls, image generation | Agents with MCP, code interpreter, web search | Built-in vector database, no-code agent builder, desktop app | | **Track record** | Launched May 2026, moving fast | Established, years of releases | Established, years of releases | Established, Docker and desktop versions | | **xCloud page** | [**Odysseus hosting**](/odysseus-hosting/) | [**Open WebUI hosting**](/openwebui-hosting/) | [**LibreChat hosting**](/librechat-hosting/) | [**AnythingLLM hosting**](/anythingllm-hosting/) | Rule of thumb: **Odysseus** suits one person or a small team that wants a whole AI workbench in one app. **Open WebUI** and **LibreChat** suit teams that mainly need dependable chat with user controls. **AnythingLLM** suits document-heavy work with the least setup. The catch worth knowing: a project that is months old will change under you. Pin a version and back up before every upgrade. ## Self-Hosting Requirements and Ways to Run It ### What You Need - A Linux server, a Mac or a Windows machine. Docker is the easiest route on all three. - Enough room for the app and ChromaDB. The xCloud template's floor is 1 CPU core, 1.5 GB RAM and 10 GB disk. - A model endpoint: a local Ollama or vLLM server, or an API key from a provider. - Outbound internet on first use, so the embedding model can download. - A GPU only if you want to serve local models on the same machine. ### Option 1: Docker Compose on Your Own Server The official quick start clones the repository, copies an example settings file and starts the stack with Docker Compose. The first admin password appears in the container logs. The trade-off is everything after the first start. You need a domain, an HTTPS certificate, a firewall, backups and a plan for updates. Keep sign-in turned on and never expose the model or database ports to the internet. **Your time is the hidden cost.** Self-hosting is cheap until the occasional 3 a.m. restart, the certificate that did not renew or the update that broke the vector store. Our comparison of [**self-managed vs managed hosting**](/self-managed-vs-managed-hosting/) walks through that decision in more detail. ### Option 2: A Managed One-Click Install A managed install handles the server, HTTPS and the container setup. You still own the app settings, the model choice and your data. ## Pros and Cons of Odysseus | Pros | Cons | | --- | --- | | ✅ Chat, agents, research, email and calendar in one app | ❌ Very young project; expect bugs and fast changes | | ✅ Free and open source (AGPL-3.0-or-later) | ❌ AGPL obligations if you modify it and offer it to others | | ✅ Works with local models and OpenAI-compatible APIs | ❌ No model bundled; you must connect or serve one | | ✅ Cookbook helps you pick models that fit your hardware | ❌ Cookbook reliability still varies across machines | | ✅ Data, memory and document indexes stay on your server | ❌ Agents with shell access need careful security settings | | ✅ Light app; runs on a small server with an API model | ❌ Shorter track record than Open WebUI, LibreChat or AnythingLLM | ## Deploy Your Own Odysseus Workspace With xCloud xCloud offers Odysseus as a one-click app on a dedicated Cloud VPS. The install pins the official **Odysseus 1.0.3** image and a **Chroma 1.5.9** vector store that only the app can reach. It creates the admin account, mounts six storage volumes and publishes the workspace over HTTPS on your domain. Getting started takes three steps: 1. **Sign up and pick Odysseus.** Create your xCloud account, choose Odysseus from the one-click catalog and set a domain. Optionally set an admin password, or leave it empty to get a generated one. 2. **Let xCloud deploy it.** xCloud starts Odysseus with its private Chroma store, creates the admin account and turns on HTTPS. 3. **Sign in and connect a model.** Open your Odysseus URL and wait for the sign-in page to finish loading before you submit. Then sign in as **admin** with the password from your installation credentials and add a model endpoint under **Settings**. A few details are worth knowing before you start: - **Self-registration is off**, and nobody can create a second admin account. Treat the admin password as full control, since agents can run commands. - **No language model or search engine is bundled.** Your model provider bills its own usage. - **Six volumes** hold accounts, uploads, logs, SSH keys, downloaded models and vector indexes. Back up all six, and treat the copy as secret, because stored API keys live there. 👉 See [**backups in xCloud**](/docs/backups-in-xcloud/). - **Upgrades are manual.** Your install stays on 1.0.3 until you choose a newer template version, so nothing changes under you. At the time of writing, the hosting page lists a **6 GB Cloud VPS (4 vCPU, 100 GB NVMe)** at $24.99 per month as the starting size. Want to limit who can even reach the sign-in page? 👉 Follow the guide to [**configure firewall management in xCloud**](/docs/how-to-configure-firewall-management-in-xcloud/). **Ready to run a private AI workspace without the setup?** [**Deploy Odysseus hosting on xCloud**](/odysseus-hosting/) **and connect your first model today.** Prefer to build it yourself from the repository? 👉 Follow the step-by-step guide to [**host Odysseus on a VPS with xCloud**](/docs/host-odysseus-on-a-vps-with-xcloud/) using the manual Docker route. ## Try Odysseus This Week and Keep Your AI Data Private So, what is Odysseus in one line? It is a private AI workbench: chat, agents, research and your daily tools in one app, connected to the model you choose. In return, you accept the pace of a project that is only months old. This week, connect Odysseus to one model you already trust and run Compare against a second. Upload a few documents, try Deep Research on a real question and decide whether it earns a place in your routine. Whichever direction you go, owning your AI data is the move worth making. If you have found this blog helpful, feel free to [**subscribe to our blogs**](/blog/) for valuable tutorials, guides, knowledge and tips on web hosting and server management. You can also join our [**Facebook community**](https://www.facebook.com/groups/xcloud.community) to share insights and engage in discussions. ## Frequently asked questions ### What is Odysseus in simple words? Odysseus is a free, open-source AI workspace you run on your own server. It gives you chat, agents, deep research, documents, email, notes and a calendar in one web app. You connect the language model yourself, either a local one or an API. ### Is Odysseus free to use? Yes, the software is free under the AGPL-3.0-or-later license. You still pay for the server it runs on and for any paid API model you connect. If you change the code and offer it to others over a network, the license asks you to share those changes. ### Is Odysseus made by PewDiePie? The first commit, titled "Odysseus v1.0", landed on May 31, 2026, from the pewdiepie-archdaemon GitHub account, and early coverage credits the YouTuber as its creator. The project now lives in the odysseus-dev organization on GitHub, with commits from hundreds of contributors. ### Does Odysseus come with an AI model? No. Odysseus is the workspace, not the model. You connect Ollama, vLLM, OpenAI or any OpenAI-compatible endpoint, and its Cookbook can help you download and serve a local model that fits your hardware. ### Should I choose Odysseus or Open WebUI? Choose Odysseus if you want chat, agents, research, email and a calendar in one app and you are comfortable with a young project. Choose Open WebUI if you mainly need a mature chat interface with detailed user roles. Both connect to Ollama and OpenAI-compatible APIs. ### I only have a small VPS with no GPU. Can I still use Odysseus? Yes. Run Odysseus on the small server and point it at an API provider or a model server elsewhere. The project's setup guide recommends exactly this for small hosts, because local model serving is the heavy part.