We built Kavim, a local-first, open-source AI brainstorming canvas for visual thinking and collaboration. It lets you chat with AI, branch ideas without losing context, and collaborate in real time — all without relying on the cloud.

Core ideas:

Visual branching that preserves context across conversations

Real-time, peer-to-peer collaboration (built on Yjs + WebRTC)

Local storage and offline-first workflows; no cloud required

Bring-your-own AI: plug in ChatGPT, Claude, Gemini, or others

Fully open source (AGPL-3.0) and free to download

No subscriptions — you only pay your AI provider if applicable

Why we built it:

We were building an AI QA tool before realizing how hard it is to think visually with AI while keeping ownership of your data. So we pivoted to create Kavim — a tool where ideas evolve like conversations, not documents.

Tech:

React + Electron + Yjs + WebRTC + OpenAI/Anthropic APIs.

Everything runs locally — no central server required.

Status:

MVP is live: https://kavim.deepelegant.com/

GitHub: https://github.com/deep-elegant/kavim

Releases (download the app): https://github.com/deep-elegant/kavim/releases

Discord: https://discord.com/invite/c99hvZMJda

We’d love your thoughts on:

Local-first architecture (Yjs + WebRTC challenges) UI/UX for branching AI chats Any use cases you’d like to see supported

Thanks for reading and testing!