Vision
Nimi starts from a simple belief: your AI should be personal. It should run where you are, remember what matters to you, and answer to you — not to a platform that owns it.
That is why Nimi is local-first. It is software you install, and the AI you talk to runs on your own machine, using models you choose, local or cloud. It is why your characters and worlds keep their identity no matter where you take them. And it is why no single part of the system quietly becomes the owner of everything else: who a character is, how it runs, and how it appears on screen are separate concerns, each with a clear home.
For users, this adds up to an AI that feels like yours. For builders, it is a stable set of boundaries to build against.
Persistent identity, local execution
The character you talk to is a persistent identity kept by Realm — the same character across every world you visit. When you start a conversation, Runtime brings that character to life on your machine as a LocalAgent, using the character description Realm provides. One character can support many local AI experiences, and each LocalAgent keeps its own running state.
User-controlled access
Apps ask for things through the SDK in typed form. Runtime works out who you are, which App is asking, and what it is allowed to do from your active sign-in session. Apps never hold your Realm credentials, model-provider keys, or any Runtime-internal proof.
Interoperable projections
Conversations, memory, knowledge, voice, and presentation all live with Runtime. Other products receive bounded, typed views of them — nothing more. Avatar can render a character richly on screen without taking over the LocalAgent behind it.
Incremental platform growth
The product loop running today does not depend on general Workflow, MCP, World Evolution, marketplace, public distribution, or commercial settlement. New capabilities can join later under their own boundaries, without redefining the core model of characters brought to life as LocalAgents.