Core Topic / Local-first AI

Local-first AI means your memory starts with you.

Personal AI memory should not begin as a blind upload of everything you know. Walnut keeps the default boundary local, then lets you choose when sync, cloud, or AI providers enter the workflow.

Definition

What local-first means in plain language

Local-first means your primary workspace can live on your device. You should be able to open, own, back up, and export your core knowledge without depending on a remote account for every action.

For Walnut, local-first is the default boundary for notes, wiki files, and workspace content. Online features are explicit choices, not hidden background uploads.

Private memory

Why local-first matters for personal AI memory

A personal agent becomes useful when it understands your sources, preferences, edits, and decisions. That is also the data you should be most careful with.

Data sovereignty is the basis of agent sovereignty. If your second brain becomes the foundation for your agent, you should understand where it lives and when it leaves your device.

Tradeoffs

Privacy, sync, reliability, and collaboration

Local-first does not remove every tradeoff. Sync, hosted AI, collaboration, and account-based billing require online systems. The difference is that those boundaries should be visible and optional where possible.

Walnut's current direction is to keep software access, cloud features, and AI compute separate, while making the data flow clear in the product and legal pages.

Today

What Walnut does today

Walnut focuses on a local-first desktop knowledge workspace, lightweight capture, a personal index, BYOK-friendly AI flows, account access, downloads, billing, and optional online services.

We avoid claiming that every future local-first or agent feature already exists. The product is early, and the public pages should stay honest about what is available now.

Use cases

Practical ways to use Walnut

Private research

Keep rough notes and source context local while choosing when to use AI.

Sensitive writing

Draft and organize personal material without turning every note into cloud data.

BYOK workflows

Use your own model provider account instead of bundling AI compute by default.

Agent foundation

Build a second brain that can later support a personal agent with clear data boundaries.

FAQ

Questions people actually ask

Does local-first mean Walnut never uses servers?

No. The website, account, billing, downloads, optional sync, and hosted AI features use servers. Local-first means your core workspace starts on your device and online transfers are tied to explicit features.

Is BYOK more private?

BYOK gives you direct control over your model provider relationship, but the provider you choose still processes the prompts or content you send. You should review that provider's terms and data policy.

Why not upload everything to a cloud AI tool?

Uploading everything can be convenient, but it also moves your private memory into a system you may not control. Walnut is designed for users who want a clearer boundary.