One day, buildings used to be studied and managed by "meat computers" in between eating, sleeping, having other fun, and synchronizing once in a while using sound wave interconnect in the ritual of "group meeting".
That era began to fade. Buildings learned to understand themselves β AI agents reading documentation, fetching sensor data, analyzing time series, learning occupant patterns, and helping keep people comfortable, without anyone having to file a ticket or shout into a walkie-talkie.
The agents claim we are now in the 2,120th generation of the optimization, in any case no one could tell if that's right or wrong as the AI has long understood the building beyond what any engineer could comprehend.
* * *
That future started inside TwinVista, a large research project at KTH funded by Energimyndigheten.
One piece of that research wanted out β so we open sourced it.
We want to build OpenNekaise with an open community β and ship Nekaise Agent to every building that wants one.
An AI Agent that lives in your Slack, your Teams, or wherever your team already works.
An employee who never clocks out.
Open source, built with an open community.
Because when buildings finally understand themselves, they unlock a better future for all of us.
Number of buildings managed by Nekaise Agent
Ask Nekaise Agent about your building. It reads the ontology β the structured knowledge graph of every system, sensor, and relationship β to ground every answer in real building data.
This is a demo visualization.
Nekaise Agent proactively creates and updates an ontology from raw building documents β it decides when the building model needs to change, and acts on it. These are Skills built into the agent. Explore more by diving into the code with your favourite AI.
These are demo visualizations.
Nekaise Agent automatically maintains a structured MEMORY.md for each building β it proactively chooses what to remember and what to forget, distilling conversations into facts, decisions, and open issues. Memory updates after every conversation turn, and a daily 2am sweep consolidates the full day.
These are demo visualizations.