Quickstart
From install to first chat turn in under 5 minutes.
1. Install Hypex
Windows one-liner:
irm https://hypex.pages.dev/install.ps1 | iexmacOS / Linux: download the tarball from the downloads page and unzip. See Install Hypex for signing notes.
2. Open a folder
Launch Hypex, then File → Open Folder… Any project works — TypeScript, Python, Rust, C, Go, whatever.
3. Pick a provider
Command palette (F1) → Hypex: Select Model Provider. Choose one:
| Provider | Best for | Needs key |
|---|---|---|
| Ollama | Local, free forever, private | No |
| Anthropic | Best reasoning (Claude) | Yes |
| OpenAI | GPT-5 / GPT-4o | Yes |
| OpenRouter | 300+ models through one key | Yes |
| xAI (Grok) | Fast reasoning, 95 % cache hit | Yes |
| Claude Code CLI | Use your Claude Max subscription | Uses CLI auth |
4. (Ollama path) pick a local model
Easiest: run Hypex: Diagnose Hardware & Recommend Local Model — we probe your RAM / CPU / VRAM and suggest a fitting model from the Gemma 5 / Gemma 3n / Qwen3-Coder / Phi-4 line. Accept the pick and Hypex kicks off ollama pull in a terminal.
Don't have Ollama installed? Hypex detects that, offers a platform-specific download link, and auto-retries the pull once you're back.
5. Start chatting
Ctrl+Shift+L focuses the chat sidebar. Type a natural-language question or:
/for slash commands (/explain,/tests,/commit,/fix)@for context mentions (@buggy.ts,@workspace,@selection)
6. Try an inline edit
Select code. Ctrl+I opens an inline prompt. Type "add JSDoc" → Hypex rewrites the selection in place. Tab accepts.
Next steps
- Agent deep-dive — how tool approvals, rewind, and checkpoints work
- Every shortcut
- Write a skill that the agent auto-applies when it matches keywords