Your private AI pod, one terminal command away.

Isolated. WireGuard-tunneled. OpenAI-compatible. Drives every AI CLI you already use — Claude Code, OpenCode, Gemini, Cursor. Yours alone.

Install

curl -fsSL https://get.traylinx.com/install.sh | bash
macOS · Linux
powershell -c "irm https://get.traylinx.com/install.ps1 | iex"
Windows (tunnel support experimental)
brew install traylinx/tap/tytus
Homebrew (macOS · Linuxbrew)

Then

tytus setup
Logs you in, picks a pod, opens the tunnel, verifies end-to-end.

What you get

One stable URL + key

Set OPENAI_BASE_URL + OPENAI_API_KEY once. Works forever across pod reallocations.

Every AI tool, instantly

tytus link drops CLAUDE.md, AGENTS.md, .mcp.json and slash commands so Claude Code, OpenCode, Gemini and Cursor drive Tytus natively.

Private by construction

Userspace WireGuard. No shared keys. Cross-pod traffic blocked at the firewall. Your conversations never leave your pod.

MCP server built in

tytus-mcp ships alongside the CLI. Point Claude Code or any MCP client at it and your model can drive Tytus on your behalf.