brains desktop

A native desktop app for working with powerful AI models — on your own subscription — wrapped in a brains‑native experience. Sign in with your brains account, pick a model (Opus, Fable, GPT/Codex…), and your memory, boards, dashboards, and automations are first‑class citizens of the chat.

---

Why we're building this

1. The existing desktop clients get in your way

Claude Desktop and the others are painful in practice:

• Fiddly installation and setup before you can do anything useful.
• A permission prompt for everything — it interrupts constantly.
• Clumsy code workflows — you have to point it at folders and hand‑wire MCP.
• The OpenAI path is even more awkward — wiring up brains is a chore, and half the time it just doesn't work well.

brains desktop is a clean, native app that just runs your model on the subscription you already pay for. Pick a genuinely capable model — Opus or Fable — and go, without the ceremony.

2. One UI, every model

Claude, Codex/OpenAI, and more behind a single consistent interface. Choose a model per session; your workflow doesn't change when the vendor does. Persistent sessions, real intermediary‑step visibility, and a native chat that feels the same regardless of who's behind it.

3. brains is native, not bolted on

• Mini‑sites and dashboards render inline in the chat — the real engine, real charts and layout, not a raw data dump.
• Tag context with @ — boards, mini‑sites, and memory entities (people, organizations, projects, places, concepts) become part of the conversation, the same way opening a dashboard does.
• Work on automations and integrations inline, right where you're already thinking — no context switch to a separate web app.

4. Features you can't get by gluing tools together

First‑party capabilities that only make sense in a native app — like recording calls straight into brains, with no third‑party bot and no export step. The desktop app is the seam where these live.

---

Status

Early and iterating fast. Forked from OpenCovibe and being rebuilt around brains.

Stack

Tauri v2 · SvelteKit (Svelte 5 runes) · TypeScript. It talks to the brains MCP server (mcp.mybrains.ai) with a per‑app OAuth token and wraps the Claude / Codex CLIs so everything runs on your subscription — no API keys required.

Develop

npm install
npm run tauri dev

Windows

The app builds and runs natively on Windows (Tauri v2 targets x86_64-pc-windows-msvc).

Prerequisites

• Visual Studio Build Tools with the Desktop development with C++ workload (MSVC linker)
• Rust with the default x86_64-pc-windows-msvc toolchain
• Node.js ≥ 20
• WebView2 runtime — preinstalled on Windows 10/11; the NSIS installer bootstraps it automatically on machines that lack it

Build

npm install
npm run tauri build

Installers land in src-tauri/target/release/bundle/:

• nsis/brains_<version>_x64-setup.exe — per-user installer (no admin required)
• msi/brains_<version>_x64_en-US.msi — MSI for machine-wide / managed deployment

Notes

• Builds are unsigned, so SmartScreen may warn on first launch of the installer — More info → Run anyway.
• Use npm install rather than npm ci: the committed lockfile currently fails npm ci under npm 11 (lockfile drift).