Install RUNE and send your first prompt
Install the RUNE CLI, pull a SAGE reasoning model, and send your first prompt directly from the terminal.
- One-line install: get the CLI running in seconds on macOS or Linux.
- Local model setup: pull a reasoning model that fits your machine — no API key needed.
- Send a prompt: have RUNE read your project and respond in the terminal.
By the end you'll have RUNE installed, a model pulled, and responses grounded in your actual codebase.
Time to complete: ~10 minutes (plus model download time)
Prerequisites
- macOS or Linux (the installer detects your platform and architecture).
- Ollama installed and running.
- 5–10 GB free disk for the CLI plus one model.
Step 1: Install the CLI
Run this in your terminal:
The installer detects your platform and architecture, downloads the latest release, and makes the rune command available, provided your PATH is configured. Restart your shell if the command isn't found right away.
Step 2: Pull a reasoning model and launch
- Pull a model that fits your machine — start with the 8B default:
Other sizes under comethrusws/sage-reasoning: :3b for constrained hardware, :14b for complex refactors, :32b for large autonomous tasks.
- Open a terminal in any project directory and run:
RUNE scans your file tree and Git status automatically — nothing runs in the cloud, and there's no sign-in or API key. Everything stays on your machine.
Step 3: Send your first prompt
Type a request directly in the prompt. For example:
List the files in this directory and explain what each one does.
RUNE reads your project context and responds with relevant information. It can also generate code, edit files, and run shell commands when you approve — try:
Map out how file uploads flow through this codebase.
Verify
You've completed the setup if:
rune --versionreturns a version number.ollama listshows your pulled model (e.g.comethrusws/sage-reasoning:8b).- RUNE responds to prompts with answers grounded in your actual files.
