The agentic AI code editor for Windows. Bring your own model — your keys never leave your PC.
Run the installer. Clean install, clean uninstall — no registry mess, no background services.
The first-run wizard validates your provider key and stores it in Windows Credential Manager.
Hand off your first task and watch the agent plan, edit, run, and verify — autonomously.
Run parallel AI agents on different tasks simultaneously. Each agent plans, codes, and iterates independently.
Set up automatic failover chains across OpenAI, DeepSeek, MiMo, OpenRouter, and Qwen. If one provider is down, the next takes over.
Connect to databases, APIs, and external tools via the Model Context Protocol. Extend your agent's capabilities.
Your API keys are stored in Windows Credential Manager. Never in plaintext, never on our servers — BYOK architecture.
Fork any agent session to explore alternate approaches. Review the full plan before code is written.
Built for Windows. Feels like a native app — not an Electron wrapper. Fast startup, low memory footprint.
The IDE is free to download. To use AI features, you need a subscription ($10/mo) which includes platform services (OCR, embeddings, web search). You also bring your own API keys for LLM providers — you pay them directly.
Your keys are stored locally in Windows Credential Manager. They never leave your machine and are never sent to any Cortex server. You pay the LLM providers directly — Cortex doesn't mark up or resell tokens.
The IDE itself works offline for code editing. AI agent features require an internet connection to reach your chosen provider's API. The $10/mo subscription covers platform services (OCR, embeddings, web search) which also require internet.
Currently Cortex is Windows-only. Mac and Linux support is on the roadmap.
Cortex is bring-your-own-key — your LLM requests go directly from your machine to the provider over TLS. Your prompts, code, and credentials never touch our infrastructure. The $10/mo subscription covers platform services (OCR, embeddings, web search) which are proxied through Cortex servers.