Alysis AI

AI lab · Athens, Greece

We build AI agents
that do real work.

Alysis is an independent AI lab. We work on the hard part of agents — getting them to carry a whole task through to a result someone can trust — and release what we learn as products people can put to work. The first of them is Alysis Code.

What we buildTalk to us

How we build.

Three commitments that decide what we ship, and what we leave alone.

The problems we work on

  • Agents should finish the job

    Not autocomplete, not another chat window. An agent should take a real task end to end and hand back something you can check.

  • Research has to ship

    A result that only holds in a demo has told us nothing. Everything we work on goes into something people use for real work, which is where we find out whether it was true.

  • No lock-in

    What we build runs on the model you choose — frontier, open-weight, or something on your own hardware. We do not make that decision for you.

What we build.

One product in the world, one way of working with companies, and the next things taking shape.

  • Alysis Code

    Live

    An open-source coding agent for the terminal. Describe a change, and it plans the work, makes it, and verifies the result before anything merges.

    Apache-2.010,000+ sign-ups3,700+ installs a month

    alysiscode.com github.com/AlysisAi/alysis-code

  • Agents for teams

    Available

    Most companies do not need another tool; they need agents that fit how they already work. We set them up, train the people using them, and handle model access.

    Enterprise seatsTrainingModel access

    How we work with teams

  • New agents

    In the lab

    Code was the first place agents became genuinely useful. We are working on the next ones, and on making any agent easy to run on any model.

    ResearchEarly access

    Talk to us about it

Latest.

Announcements and notes from the lab.

All news