# golutra

> Keep the coding CLI you already use — golutra runs several of them at once as a squad you can watch

- Category: AI Coding / Coding-Agent Harnesses & Runtimes
- Language: Rust
- Licence: BSL-1.1
- Repository: https://github.com/golutra/golutra
- Website: https://www.golutra.com/
- Canonical page: https://ai-tldr.dev/tools/golutra/

## In short

golutra is a desktop multi-agent workspace built around the coding CLIs you already have installed. It sits in the Coding-Agent Harnesses & Runtimes category of the AI Coding tools and is open-source, written in Rust and licensed BSL-1.1. To get started: install the CLIs you want it to drive first.

## Overview

golutra is a desktop multi-agent workspace built around the coding CLIs you already have installed. Rather than replacing Claude Code, Gemini CLI, Codex, OpenCode, Qwen Code or OpenClaw with another agent, it wraps them: the same commands keep working, the same projects stay where they are, and golutra adds the layer above — running several of them in parallel, scheduling and tracking their work, and handing results from one to the next.

The interface is a visual one over live terminals. Each agent appears as an avatar you can click to inspect its log; you can inject a prompt directly into a running terminal stream to correct an agent mid-task, or leave the squad working in the background and watch the status board. That combination — a background terminal you can still reach into, with the project context attached — is what the project calls its stealth terminal, and it is the answer to the usual alternative of tiling half a dozen terminal panes by hand.

It is aimed at long-running collaboration rather than a single interactive session. Workflows are user-defined and can be exported and imported as templates in one click, which is how the same installation ends up driving quite different jobs — a software team's build-and-review loop, a one-person company's pipeline, automated long-form writing, or video and social-content production. Two companion projects extend it: `golutra-mcp` connects external tools and agents through `golutra-cli`, and EverOS can act as a memory layer that carries context and project knowledge across sessions.

The application is a Tauri desktop app with a Vue 3 front end and a Rust core, released for Windows, macOS and Linux. It is written and maintained by a single developer, `seekskyworld`, and the source is published under the Business Source License 1.1: building commercial software with golutra is allowed and whatever you produce with it is yours, but the licence is source-available rather than OSI open source.

## What it does

- Drives existing coding CLIs unchanged — Claude Code, Gemini CLI, Codex CLI, OpenCode, Qwen Code, OpenClaw, or any other CLI — with no project migration and no new commands to learn
- Runs agents in parallel with automatic result handoff, so several tasks progress at once instead of one terminal at a time
- Per-agent logs behind each avatar, and direct prompt injection into a live terminal stream for instant course correction
- One orchestration layer tracking status and scheduling across every CLI, with test, build and regression output aggregated into a single delivery path
- Session-level context and prompt reuse, so requirements do not have to be restated for each agent
- Custom workflows with one-click template import and export, built for long-running automation rather than short sessions
- Optional memory layer through EverOS, and MCP connectivity through the companion `golutra-mcp` project
- Native desktop builds for Windows, macOS and Linux (Tauri, Vue 3 and Rust)

## Getting started

golutra is a desktop application, distributed as signed installers on its GitHub releases page. The repository itself holds the source and the releases; there is no package-manager install.

1. Install the CLIs you want it to drive first

   golutra orchestrates agents that are already on your machine, so install and sign in to at least one of them before starting. Any CLI works; these are the ones the project lists explicitly.

```bash
# whichever you already use — e.g.
claude --version
codex --version
gemini --version
```

2. Download the desktop build

   Take the installer for your platform from the releases page, or the download link on the project site.

```text
# https://github.com/golutra/golutra/releases
# Windows · macOS · Linux
```

3. Point it at a project and add agents

   Open a project directory, then add agents to it — each one is a CLI golutra will launch in its own terminal. Because nothing is migrated, the project is the same working copy your CLI would see on its own.

4. Run them in parallel and steer from the board

   Start the squad and let the agents work in the background. Click an avatar to read that agent's log, and inject a prompt straight into its terminal stream when it needs redirecting rather than restarting the task.

5. Save the loop as a workflow template

   Once a sequence works — say analysis, implementation, test, review — export it as a workflow template so the same orchestration can be re-imported or shared instead of rebuilt by hand.

## When to use it

- Reach for it when you are running three or four coding CLIs in tiled terminals and losing track of which one is waiting on you
- Reach for it to keep agents working on a long task in the background while you watch status and step in only when a log looks wrong
- Reach for it when a repeatable multi-step loop — analyse, implement, test, review — is worth saving as a template rather than re-typed each time
- Reach for it if you want a visual command centre but are not willing to give up the CLI you already trust