Utopia

[POkorji]

Y'all building the community?

[WaylandYang]

Working on it. For now everything's here on GitHub: Discussions, Issues, and CONTRIBUTING.md if you want to pitch in.

Discord is coming once there are enough of us to fill it. Thanks for the interest!

[WaylandYang]

Thank you, that's very kind. We're not set up to take donations currently, but if you'd like to help, running it and opening issues would mean a lot to us right now :)

[POkorji]

I feel you. The setup for the donation is very easy and straightforward.

Utopia has a lot to offer

[Sm-bello]

Honestly what's utopia in simple lay man terms? I'm looking at building my own local ai agent that works on my laptop offline, and both online, can this tech be a support to that goal?

[WaylandYang]

Good question. The README is still a work in progress and we keep improving it; questions like this tell us what's missing. Here's the short version.

Utopia builds a working model of an organization from what it already writes down. Point it at the documents, tickets, pages and data an organization produces, and it keeps track of the people, products, contracts and events in them, how they relate, and what was true when. Every fact keeps its source and its history, so you can ask what changed, since when, and why.

On top of that model it reasons. The rules live in the ontology, and the engine uses them both ways: to derive facts nobody wrote down, each with a proof chain back to the sentences it rests on, and to flag contradictions between what the documents say and what the rules allow. A person decides what the model is allowed to conclude; nothing derived can overrule something asserted.

The next step, and the reason for all of the above, is controlled execution: an agent's actions checked against the same rules before they run, and a decision replayable afterwards with the grounds it stood on at the time. That part is on the roadmap, not shipped.

It is built for teams and companies, or for a group of agents that has to share one governed picture of the world instead of each keeping its own. For a single personal agent on a laptop it is more than you need. It does run fully local (one binary plus Postgres, any OpenAI-compatible model endpoint such as Ollama, online or not), but its design is about shared knowledge and the decisions made on it, not one agent's private memory.

Where it could fit your project: if your agent has to work over a body of documents — a company wiki, a project's tickets, a research archive — or if it grows into several agents that must agree on the same facts, Utopia can be that shared layer. Your agent connects over MCP to search, look up entities, read facts as of a date, and see what changed. If that's the shape of what you're building, try the Quick start and tell me where it gets in the way. Thanks for asking, and I hope this helps.

[Sm-bello]

Ok this is good, I'm interested. Waiting for you to ship