IntroducingRUNE.
A local-first command-line coding assistant powered by SAGE models. Use natural language to explore, modify, and interact with your codebase—entirely on your own machine.
Get started with SAGE models.
Purpose-built reasoning models that run entirely on your hardware. Pick the size that fits your machine.
Default
8Bsage-reasoning:8b
Balanced performance for everyday coding tasks. Runs comfortably on most modern laptops.
Pull modelInstinctive
3Bsage-reasoning:3b
Ultra-fast responses for lightweight tasks. Designed for constrained environments and quick iteration.
Pull modelIntuitive
14Bsage-reasoning:14b
Deeper reasoning for complex refactors, multi-file edits, and nuanced architectural questions.
Pull modelAgentic
32Bsage-reasoning:32b
Maximum reasoning depth for large-scale autonomous tasks. Orchestrates subagents, plans across entire codebases, and handles long-horizon engineering problems.
Your code stays on your machine.
Rune runs SAGE reasoning models locally through Ollama. Every file read, every code edit, every shell command executes on your hardware. Nothing is sent to external servers.
It automatically scans your project's file structure and Git status, giving the agent deep contextual awareness of your codebase without any manual setup.
One command to get started.
The installer detects your platform and architecture, downloads the latest release, and configures everything automatically.
$ curl -fsSL https://sagea.space/rune-cli/install.sh | bashA complete coding toolkit.
Rune comes with a powerful set of tools accessible through natural language conversation.
File operations
Read, write, and patch files with precision. Apply targeted edits using search-and-replace without rewriting entire files.
Shell execution
Execute commands directly in a stateful terminal session. Run tests, build projects, and manage dependencies.
Code search
Recursively search your codebase with ripgrep support. Find patterns, definitions, and usages across thousands of files.
Subagent delegation
Delegate tasks to specialized subagents that work independently, preventing context overload on complex projects.
Interactive questions
The agent asks clarifying questions with structured options when it needs more information, enabling collaborative workflows.
Skills & extensions
Create custom slash commands and reusable skill definitions. Share workflows across teams and projects.
Multiple agent profiles.
Choose the right agent for the job. Each profile has different safety and automation defaults, so you stay in control.
defaultStandard agent. Requires approval for all tool executions.
planRead-only agent for exploration and planning. Auto-approves safe tools.
accept-editsAuto-approves file edits only. Useful for focused code refactoring.
auto-approveAuto-approves all tool executions. Use with caution.
Questions, answered.
Everything you need to know before installing Rune.
Yes. Rune is designed for local-first workflows. It runs on your machine and works with local model runtimes so your source code and prompts remain under your control.
Build your own AI future
Build, customize, and deploy AI solutions with complete control.