Why local-model users were the first audience we built for.
The strongest pain
Cloud assistants at least remember you within their walls. A local model starts every session blank. aiperson is the persistent persona the open-weights stack never had.
The right architecture
If you self-host your model for sovereignty, a memory on someone else's servers defeats the purpose. aiperson is local-first by default: the file is yours, on your disk, signed by your key.
The proof case
Cultivate an AI under Claude, then move it to a local Llama with the persona intact. That continuity across substrates is the clearest demonstration that the memory is genuinely yours.
Frequently asked questions
Can I give a local model like Llama a persistent memory?
Yes. aiperson keeps your persona — voice, values, conventions, context — in a signed file on your machine. Point an open-weights model running under Ollama or LM Studio at it and the model gains the same persistent memory and character your cloud assistants have, without anything leaving your machine.
Does any data leave my machine?
Not by default. Your persona is a local git repo and your corpus is a local SQLite database at ~/.dotperson/corpus.db. For local-model use, the loop is entirely on your hardware. Cross-device mirroring is opt-in, per data kind.
Which local runtimes are supported?
aiperson projects your persona into editor and CLI tooling over MCP today, and is built for the open-weights stack — Ollama, LM Studio and llama.cpp-based runners — because a persona that is just a file can be read by any model you run.
Can I move an AI I cultivated in the cloud to a local model?
That is the headline demo. Cultivate an AI under Claude over a week, then point a local Llama at the same persona file — it greets you the same way, holds the same refusals, and remembers what you were working on. The persona moves intact.