The 30-second version
On May 31, 2026, Felix Kjellberg — PewDiePie — shipped Odysseus, an open-source AI workspace you run entirely on your own machine. It crossed 30,000 GitHub stars in its first 48 hours, and for good reason: it's a clean, self-hosted home for chat, autonomous agents, research, email and notes that can run open-weight models on your own hardware.
13x shares Odysseus's core conviction — your machine, your data, no telemetry — but aims it at a narrower target: developers running coding agents. Where Odysseus is a general AI workspace, 13x is a control layer for the agents that write your code.
If you want one line: pick Odysseus to own your general-purpose AI; pick 13x to command a grid of coding agents on a real codebase. Plenty of people will happily run both.
What Odysseus is
Odysseus is a self-hosted workspace that puts a single interface in front of your language models. Out of the box it covers multi-turn chat, autonomous agents that plan and call terminal tools — bash, file edits, web browsing — with persistent memory, multi-step web research that synthesises sources into reports, and productivity surfaces like email triage, notes and calendar.
Its defining trait is that it can run the model too. Odysseus supports local inference engines — Ollama, vLLM and llama.cpp — alongside cloud APIs like OpenRouter and OpenAI. It scans your CPU and GPU, scores your hardware, and offers one-click serving across a catalogue of 270+ open-weight models. The pitch is privacy maximalism: the interface and, if you want, the model itself never leave your machine.
- Free and open-source, with a fast-moving community.
- Genuinely local inference — run open-weight models on your own GPU, no API key required.
- Broad surface area: chat, agents, research, email, notes and calendar in one place.
What 13x is
13x is a local-first control workspace for developers who run coding agents. Instead of one chat window, it gives you a grid of up to twelve real terminal shells, all rooted in the same project, so you can run several agents — Claude Code, Codex and friends — in parallel and actually keep track of them.
Around that grid sit the features that make many agents manageable: named rigs that save and relaunch a whole workspace layout in a click; a codebase map that scans your repo into a local symbol graph so you can check a change's blast radius before you make it; context packets that turn a task sentence into a budgeted, local-only payload so you send agents less noise; and pulse, which surfaces which session needs you right now.
- Built for parallelism — a slot per agent, not one session at a time.
- Codebase-aware: a local symbol graph, blast-radius checks and context packets.
- Telemetry-free and local-first, like Odysseus — your code never leaves the machine.
The shared DNA
It's worth dwelling on how much these two agree. Both are local-first. Both are telemetry-free. Both reject the idea that using AI well requires shipping your data to someone else's cloud. In a market full of editors that quietly index your repository server-side, that shared stance is the headline, not a footnote.
The launch of Odysseus matters to anyone building in this space — us included. A creator with PewDiePie's reach making "self-hosted, no telemetry" a mainstream expectation is good for every tool that already believed it.
Where they diverge
The difference is the job, not the values. Odysseus is horizontal: one workspace for whatever you point a language model at, from drafting email to researching a topic to chatting with a local model. 13x is vertical: one workspace for the specific, messy job of running coding agents against a codebase and shipping the result.
That shows up everywhere. Odysseus runs the model and gives an agent a single terminal; 13x assumes you already have capable coding agents — often the frontier CLIs — and gives you twelve terminals, a map of your code and a way to orchestrate them. Odysseus optimises for breadth and privacy of inference; 13x optimises for throughput and control when more than one agent is editing the same repo.
Side by side
| Odysseus | 13x | |
|---|---|---|
| Primary job | General self-hosted AI workspace | Control layer for coding agents |
| Runs the model itself | Yes — Ollama, vLLM, llama.cpp | No — drives your coding agents |
| Parallel agents | Single agent + tools | Grid of up to 12 terminals |
| Codebase map / context packets | — | Local symbol graph, blast radius, packets |
| Saved workspace layouts | — | Named rigs |
| Session monitoring | — | Pulse |
| Local-first / telemetry-free | Yes | Yes |
| Licence | Open-source, free | Proprietary · Free / €49·mo / from €99 lifetime |
| Platforms | Cross-platform (self-hosted) | Windows 10/11 today |
Compiled from Odysseus's public launch materials and the 13x feature set — corrections welcome.
Where Odysseus wins
Be honest about it: there are real things Odysseus does that 13x does not. It's free and open-source, so you can read every line and self-host it forever. It runs the model itself, which 13x doesn't — if your goal is to never depend on a hosted API, Odysseus gets you there. It's cross-platform today, where 13x ships for Windows 10 and 11 first. And its scope is far wider than code.
If your need is "a private, self-hosted home for AI across my whole life," Odysseus is the more natural fit, and we'd tell you so.
Where 13x wins
13x earns its place the moment you're running more than one coding agent on a real repository. A single chat-plus-terminal loop doesn't help when three agents are touching the same files; a grid with saved rigs, a codebase map and pulse does.
- Running 3+ agents on one repo at once — a slot each, not one window at a time.
- Checking what a rename breaks before you ship it — local blast-radius, offline.
- Sending agents less context — budgeted, local-only packets instead of the whole tree.
- Rebuilding the same setup every morning — save it once as a rig, relaunch in a click.
They're complementary, not rivals
This isn't really a cage match. The cleanest setup for a lot of developers is both: run Odysseus as your private, general-purpose AI workspace, and run 13x when it's time to point coding agents at a codebase and ship. They share an ethos, so they sit together comfortably — neither asks you to send your data anywhere.
Which should you choose?
Choose Odysseus if you want a free, open-source, self-hosted workspace for AI across email, research, notes and chat, and you want to run open-weight models on your own hardware.
Choose 13x if you're a developer who runs coding agents and you've outgrown one-session-at-a-time — you want a grid, named rigs, a codebase map and pulse, all local and telemetry-free.
And if you're both a privacy-minded generalist and a developer drowning in agent tabs? Run both. That's the point of a local-first world: the tools are yours to combine.