CortexOS
Self-Hosted AI Operating System
Privacy-first, local-first. No telemetry. No cloud dependency.
Run everything on your own hardware.
curl -fsSL https://cortexcloud.online/install.sh | bash
The AI OS that runs on YOUR hardware
CortexOS is a fully open-source, privacy-first AI operating system designed to run entirely on your own hardware. Built for developers, researchers, and teams who refuse to send sensitive data to third-party cloud providers — CortexOS brings enterprise-grade AI capabilities behind your firewall, under your control.
Local-first inference is at the heart of CortexOS. It natively integrates with Ollama, vLLM, and llama.cpp — giving you full flexibility to choose your inference backend. Run Llama 3, Mistral, Gemma, Phi, DeepSeek, and dozens of other open models locally with zero cold-start overhead.
CortexOS ships in two modes: Hybrid Offline for fully air-gapped environments, and Full Stack for users who want the complete CortexCloud experience on their own server. When local models aren't available or you need broader coverage, an optional OpenRouter fallback routes requests through your configured API key — always optional, always transparent.
Hybrid Client-Server
React PWA frontend communicates with a local FastAPI backend. All data stays in your SQLite database. Vector memory is stored on-device. No outbound connections unless you explicitly configure them.
CortexOS is a Progressive Web App — install it from your browser on any device. Works offline for cached conversations. No app store required.
Everything you need, nothing you don't
CortexOS ships as a collection of self-contained modules. Enable what you need. Disable what you don't. Every module runs entirely on-prem.
Chat Interface
Branching conversations with full history, model switching mid-thread, and real-time streaming responses. Context-aware and lightning fast.
Autonomous Agents
Multi-step tool-using agents with planning, persistent memory, and self-correction loops. Agents that actually finish the job.
Agent Cookbook
50+ pre-built agent recipes. Deploy research assistants, writing agents, and triage bots instantly with one click — no config required.
Deep Research
Multi-hop reasoning chains that crawl the web, synthesize findings, and cite sources autonomously. Hours of research in minutes.
Documents + Copilot
Rich text editor with inline AI powers: summarize, draft from scratch, rewrite tone, translate language — all without leaving the doc.
Memory & Skills
Persistent vector memory per user and per agent. Reusable skill modules shared across workflows. Agents that remember and improve.
Email Integration
SMTP/IMAP support. Agents can read, draft, send, and triage your email automatically. Smart labeling and priority inbox built in.
Calendar (CalDAV)
CalDAV sync with any calendar server. Let AI schedule meetings, block focus time, manage recurring events, and send invites on your behalf.
Tasks & Projects
Kanban-style task management with AI-powered prioritization, deadline prediction, and intelligent delegation across your team or agents.
Scheduled Automations
Natural language cron. Just say "Every Monday at 9am, send the weekly digest." CortexOS handles the scheduling, execution, and retries.
RL Training
Fine-tune models on your private data using RLHF — entirely on-prem. No data leaves your infrastructure. Improve models on your specific domain.
Themes
Cyberpunk neon, Jank Mode, Dark Pro, and more. The UI is fully customizable. Because enterprise software doesn't have to look like 1998.
Privacy Controls
No telemetry, no analytics, no external calls. Air-gap compatible. Every network call is logged and configurable. You own the data. Full stop.
Docker Deploy
One-command Docker Compose setup. Or run bare-metal with ./start.sh. Works on Linux, macOS, and Windows WSL2 out of the box.
Modern stack, boring in the best way
No exotic dependencies. No lock-in. Technologies your team already knows, assembled into something exceptional.
The full stack is built for longevity. React 18 with TypeScript gives you a battle-tested frontend. FastAPI delivers blazing-fast async Python APIs. SQLite keeps the storage layer simple, embedded, and zero-maintenance — with a vector extension for semantic search.
Every component is replaceable. Don't want SQLite? Swap in Postgres. Prefer another inference backend? Plug it in. CortexOS is modular by design.
Up and running in 60 seconds
Pick your preferred install method. All paths lead to the same result: CortexOS running on your machine in under a minute.
curl -fsSL https://cortexcloud.online/install.sh | bash
git clone https://github.com/jonahthan433/CortexOS cd CortexOS docker compose up -d
git clone https://github.com/jonahthan433/CortexOS cd CortexOS ./start.sh
Join the Community
CortexOS is built in public. Every feature, bug, and architectural decision happens in the open. Come build with us.
GitHub
Star the repo, fork it, open issues, and submit pull requests. The entire project — frontend, backend, docs — lives in one open repository.
View Repository →Discord
Join the CortexOS Discord to get help, share what you're building, discuss new features, and connect with other self-hosters and AI builders.
Join Server →Contribute
Submit a new module, fix a bug, improve the docs, or translate the UI. Contributions of all sizes are welcome. Check the contributing guide to get started.
Contributing Guide →Enterprise on top of CortexOS
CortexCloud Enterprise takes CortexOS to production scale. Add SSO, RBAC, multi-tenant isolation, audit logging, SLA-backed support, and a dedicated deployment team — all running on the same open-source core you already control. Zero lock-in. Just more capability, when you need it.
CortexOS is always free. Always MIT. Enterprise features are additive, never paywalled from the open-source core.