Odysseus

Odysseus is an open-source AI workspace by PewDiePie that runs on your own computer or server. It is pitched as a local alternative to separate ChatGPT and Claude windows: model chat, agents, tools, memory, documents, email, tasks, and source research live in one interface.

What it does

• chats with local and cloud models through vLLM, llama.cpp, Ollama, OpenRouter, OpenAI, and GitHub Copilot
• runs an agent with access to files, web search, shell, memory, skills, and MCP
• uses Cookbook to recommend, download, and serve local models based on your hardware
• runs multi-step source research and returns a written Deep Research report
• helps with email, notes, tasks, calendar, documents, and side-by-side model comparison

Who it is for

Odysseus is worth a look if you want to keep your personal AI layer on your own machine: developers, researchers, writers, and people who do not want every file, email, and memory handed to a hosted service by default. If you only need local model serving, start with Ollama or LM Studio. If you need one API for many hosted models, compare it with OpenRouter.

What to keep in mind

The project is young and changing quickly. Treat it as an ambitious open-source workspace, not a polished enterprise product. Setup requires Docker or Python, and the agent can touch powerful surfaces such as files, shell, web search, and tokens. Run it locally, avoid direct public exposure, and review access settings carefully.

The project is available on GitHub. The official site says Odysseus is local-first, privacy-first, and has no telemetry. For a related local workspace idea, also see AnythingLLM.