Open source · Cerebral OS
Control AI agents with your mind using EEG headsets
A desktop app for thought-assisted AI agent orchestration — local-first, IDE-style, with your models and your keys. Optional EMOTIV / Cortex integration when you want neural signals in the loop.
Licensed under AGPL-3.0. No bundled cloud backend: connect Ollama, LM Studio, or your own API keys.
What it is
Cerebral OS is a standalone Electron app: chat with agents, run a real terminal against your repo, wire up skills, and optionally bring in neural signals from supported headsets. It is built for people who want automation and coding assistance without surrendering control of where inference runs.
Stack: Electron · Vite · React · TypeScript · SQLite (better-sqlite3) with secure credential storage. Bundle id com.cerebral.os.
Highlights
Agents & chat
Multi-session agent chat, composer modes (vibe / imagine / execute), tools and commands with approvals where you configure them.
Model providers
Ollama, LM Studio, llama.cpp servers, OpenRouter, OpenAI, Anthropic, Gemini, custom OpenAI-compatible endpoints, and local GGUF flows.
Workspace
Open a project folder, browse files, integrated terminal (node-pty), editor tabs, and workspace-scoped actions.
Security
API keys in the OS-backed secure store. You choose providers and can enable local-only mode so cloud routes stay off.
Optional depth
EMOTIV / Cortex headset hooks, skill marketplace, swarm UI, and in-app encyclopedia — enable what you need in settings.
Run from source
Node.js 20+, npm, and platform build tools for native modules (see the README for Windows / Linux toolchain notes).
git clone https://github.com/Theinventor17/cerebral-os.git cd cerebral-os npm install npm run dev
The Electron shell opens automatically — this is a desktop app, not a generic browser tab. Full install, build, and provider tables live in the README on GitHub.
Cerebral OS can run terminals and touch your files — only open workspaces you trust. Report security issues responsibly via GitHub when you file advisories or add a SECURITY.md to the repo.