Skip to main content

Intro

Get started with Rune.

Rune is a fast, keyboard-driven IDE for power users. The Unix way, finished as a product: code, terminals, CLI tools, language intelligence, debugging, and AI agents, all in one composable, multi-workspace environment.

Prerequisites​

Rune runs on macOS and Linux. Find your platform below for the specifics.

macOS​

  • Architecture: Apple Silicon or Intel.
  • Version: macOS 13.3 (Ventura) or newer.

Linux​

  • Architecture: x86_64 or arm64.
  • glibc: version 2.36 or newer (Debian 12+, Ubuntu 22.04+, Fedora 37+, or RHEL/Rocky/AlmaLinux 10+). The requirement is a floor, not a ceiling, so newer versions are fine. musl-based distributions such as Alpine are not supported.
  • C++ runtime: a recent system C++ standard library (libstdc++ from GCC 12 or newer). Every desktop install that meets the glibc requirement already includes it; only minimal or headless systems may need to add it.
  • Graphics: a graphical environment (X11, or Wayland via XWayland) with the system OpenGL and X11 libraries installed and an OpenGL-capable driver. On a minimal or headless install you may need to add your distribution's OpenGL (Mesa) and X11 client library packages. Most desktop installs already include them.

Install​

curl -fsSL https://api.rune.build/install.sh | sh

FAQ​

Short answers to the things we get asked most. Missing something? Get in touch.

Rune​

What is Rune?​

Rune is a fast, Unix-inspired IDE that ends the classic editor tradeoff between deep IDE integration and fast keyboard-driven workflows. Rune gives you everything a programmer needs: language intelligence, a debugger, terminals, a CLI-friendly environment, and provider-agnostic AI agents in one composable, multi-workspace environment.

Why a new editor? Why not extend other editors like VSCode?​

Because IDEs got the abstraction backwards. A GUI is a terrible integration layer: every tool becomes a panel, every workflow a menu. The terminal has been the right integration layer for fifty years: composable, scriptable, language-agnostic. Rune builds the IDE on top of it, instead of trying to swallow it. And yes, you can run Vim inside Rune. That's the whole point.

Who is Rune for?​

Rune is for programmers who live in the terminal and want production-grade IDE tools without IDE fatigue: language intelligence, debugging, AST-level search, and provider-agnostic AI agents. None of the busy panels, plugin sprawl, and mouse errands of mainstream IDEs. None of the arcane configs and brittle plugin setups of traditional terminal editors.

How is Rune different from Cursor, Windsurf or Antigravity?​

These IDEs are all VSCode forks built around their own agents. Rune is built to run agent CLIs (Claude Code, Pi, OpenCode, or your own harness) as first-class terminal clients, and includes its own agent when you want batteries included. Your harness. Your model. Your call.

How is Rune different from Claude Code, Codex or OpenCode?​

Claude Code, Codex and OpenCode are agent harnesses that grep their way through codebases, burning turns and tokens to approximate what a real IDE answers in microseconds. Rune lets agents query by symbol name and get fuzzy-completed definitions, references, implementations, docs, and debugger context. Built into our agent. Exposed to yours via skills and MCP.

How is Rune different from Zed?​

Zed is what VSCode should have been: native, fast, no Electron, no bloat. Rune isn't trying to be a better VSCode; it's built around a completely different UX: a workspace of panes you can split, rearrange, and reshape on the fly, where the terminal is the integration layer and editors, agents, and CLI tools all live under the same keybindings. Different category.

What platforms does Rune support?​

MacOS (Apple Silicon and Intel) and Linux (arm64 and x86_64). Linux support is currently in beta. Rune does not support Windows.

Can I use VSCode extensions with Rune?​

No. VSCode extensions are a JavaScript plugin system for a GUI editor, the abstraction Rune is moving past. Rune is Unix-extensible: if it runs in a terminal, it's already a plugin. Bind it to a key and you're done. For tighter integration, the runectl CLI exposes Rune's IDE APIs: editor commands, LSP intelligence, AST-level search, window management, workspace execution, storage, notifications, and MCP, to your shell, scripts, and tools. For deeper native extensions, use Rune's SDKs.

What languages does Rune support?​

300+ languages with syntactical support: highlighting, AST search, navigation. Deeper semantic support and a debugger are available for a curated set, starting with Go and expanding from there. See the Supported Languages page for the full list.

How do I run multiple agents without tmux?​

You don't need tmux: Rune's window manager is the multiplexer. Spawn each agent as its own pane or tab, split and rearrange with the keyboard, switch workspaces, and watch workspace tabs light orange when one needs you. Unlike tmux, Rune isn't stuck inside a teletype. tmux is excellent, but the terminal protocol underneath dates from the '70s: <ctrl-shift-letter> collapses into plain <ctrl-letter>, Alt is encoded as an Escape prefix that races your real Escape key, and Super/Cmd doesn't exist at all. Bind a tmux prefix, a shell key, and an editor leader, and you'll spend an afternoon avoiding collisions. Rune owns the input layer end-to-end: Meta, Alt, Ctrl, Shift, and any combination work natively, configured in one place: window manager, editor, and panes share the same key map.

Can I run other tools (Vim, Claude Code) inside Rune?​

Yes. If it runs in a terminal, it runs in Rune as a first-class citizen. Run it in a terminal, turn it into a task that reloads on file changes, bind it to a key for one-shot execution, or alias a complicated command-plus-arguments into a single named command. Rune hosts your stack instead of competing with it.

What are the release channels?​

See rune.build/releases.

AI & Models​

Where does inference run?​

Wherever you point it. Bring your own API key for Anthropic, OpenAI, or Google, or run a local model via llama.cpp. Rune does not host or proxy inference.

Can I use my existing ChatGPT Pro or Claude subscription?​

Yes, via the "models providers" Rune shell command: OAuth into your existing subscription instead of providing an API key.

Does Rune train models on my code?​

No. We don't train models. Whatever the model provider you chose does with your prompts is governed by their policy.

Can I run Rune fully offline?​

Yes. If you need an agent, point Rune's built-in agent at a local model running on llama.cpp and you're offline end-to-end.

Data & Privacy​

Where does my code go?​

Nowhere we run. Agents execute locally on your machine. Source code is never sent to Rune's servers.

What data does Rune send to its own servers?​

Two things, and only when you consent:

  • Account info: email, name, plan. Stored in our identity provider and account database.
  • Crash reports: uploaded only if you opt in, used for debugging.

What about model traffic, prompts and responses?​

Routed directly from the Rune client to the provider you chose. Rune is not in the path; the provider's privacy policy governs what they do with it.

Do you persist my code, prompts, or model responses?​

No.

Where can I find your privacy policy, terms, and EULA?​

See Privacy, Terms, and Support Policy.

Company​

Who builds Rune?​

Unstable Build, founded by Ernest Romero and Atzarí Surinyach. Rune is our single product.

Where is Unstable Build based?​

Unstable Build is a US Limited Liability Corporation. The team works remotely.

How do you make money?​

Rune subscriptions and enterprise deals. No ads, no data sales, no per-token surcharges.

Are you hiring?​

Not actively. Follow our LinkedIn jobs page to see future openings.

Ask Rune Agent