SkillAgentSearch skills...

rove

Rove — the agent multiplexer for your terminal. Run coding agents on parallel tasks with isolated worktrees and persistent sessions.

Install / Use

npx skills add Sma1lboy/rove

Installs into whichever agent you are using.

About this skill
📦

Other

Other agent config

Quality Score

88/100

Supported Platforms

Claude Code
GitHub Copilot
OpenAI Codex

Our assessment of rove

rove scores 88/100 on our quality scale, 293rd of 814 AI & Machine Learning skills we index (top 36%).

Its Other is 11 KB long, well organised into 12 sections with 8 code examples: a thorough specification that gives an agent plenty to work with.

It has 147 GitHub stars, a meaningful sign that others use it.

Substance
29/30
Structure
20/20
Description
15/15
Adoption
9/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated today, so rove is actively maintained.
  • Our last check on 2026-09-28 found the source still online.
  • It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-09-24. Automated pattern scan on 2026-09-24. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

rove compared with similar skills

All 4 of these similar skills score higher than rove; compare them before choosing.

SkillScoreStarsUpdatedFormat
rove (this skill)by Sma1lboy88147todayOther
claude-memby thedotmack10094.8ktodayCLAUDE.md
Agent-Reachby Panniantong10085.9k12d agoCLAUDE.md
Understand-Anythingby Egonex-AI10084.4ktodayCLAUDE.md
headroomby headroomlabs-ai10074.0k1d agoCLAUDE.md

Frequently asked questions

How do I install rove?
Run npx skills add Sma1lboy/rove. The install tabs above show the steps for each supported agent.
Which AI agents does rove work with?
It is written for Claude Code, GitHub Copilot and OpenAI Codex, as a Other file. Other agents that read the same format can often use it too.
Is rove safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is rove still maintained?
The repository was last updated today, so rove is actively maintained.

Rove — the agent multiplexer in your shell

<p align="center"> <img src="docs/assets/brand/bracket-chip.gif" alt="Rove — the agent multiplexer in your shell" /> </p> <p align="center"> <strong>Multiplex your agents like you multiplex your terminals.</strong><br /> Rove is an open-source agent multiplexer: it runs <a href="https://claude.com/claude-code">Claude Code</a>, <a href="https://github.com/openai/codex">Codex</a>, <a href="https://github.com/github/copilot-cli">Copilot</a>, Kimi — or any CLI you register — in parallel,<br /> each session on its own git worktree and branch — attach, detach, reattach; they keep working after you disconnect. </p> <p align="center"> <a href="https://www.npmjs.com/package/@sma1lboy/rove"><img src="https://img.shields.io/npm/v/%40sma1lboy%2Frove?style=flat-square&label=npm&color=c96442" alt="npm version" /></a> <a href="https://github.com/Sma1lboy/kobe/actions/workflows/ci.yml"><img src="https://img.shields.io/github/actions/workflow/status/Sma1lboy/kobe/ci.yml?branch=main&style=flat-square" alt="build" /></a> <a href="./LICENSE"><img src="https://img.shields.io/badge/license-MIT-blue?style=flat-square" alt="MIT license" /></a> </p> <p align="center"> <a href="https://docs.kobe.sma1lboy.me"><strong>Documentation</strong></a> · <a href="https://docs.kobe.sma1lboy.me/docs/quick-start">Quick start</a> · <a href="https://docs.kobe.sma1lboy.me/docs/concepts">Concepts</a> · <a href="https://docs.kobe.sma1lboy.me/docs/cli">CLI</a> · <a href="https://docs.kobe.sma1lboy.me/docs/api">Agent API</a> · <a href="https://kobe.sma1lboy.me">Website</a> </p> <p align="center"> <img src="docs/assets/workspace.png" alt="Rove workspace — task sidebar, the embedded engine session mid-review, and the worktree's file tree" /> </p>

Terminal multiplexers let one terminal hold many shells that survive you. Rove does the same for AI coding agents: one TUI holds many engine sessions, each isolated on its own git worktree and branch, all alive after you close the laptop lid on your SSH connection. Start the auth refactor, start the flaky-test hunt beside it, walk away, come back to two finished branches. It runs where your code already lives — laptop, devbox, VPS — with no desktop app and no browser required.

Why Rove

  • Safe parallelism — every task is git worktree + engine session + branch. Agents never trample each other or your checkout.
  • Sessions survive you — quit the TUI, drop SSH, restart the daemon; reattach and the screen comes back. A separate PTY host owns the sessions, so nothing you close kills them.
  • Real engines, real environment — Rove embeds the actual interactive CLIs next to your dependencies, services, and credentials. No API wrappers, no re-rendered streams.
  • Any engine, mid-task — claude, codex, copilot, kimi, or your own command. Hand a stuck conversation to another vendor without losing its context.
  • Terminal first — notifications and clipboard ride SSH back to your local terminal, and the whole UI folds down to a phone-width session.
  • Agents orchestrating agents — rove api lets a script, or another AI agent, spawn tasks, supervise them, and land the results headlessly.
<p align="center"> <img src="docs/assets/demo.gif" alt="Rove demo — two tasks running at once, each on its own worktree and branch" /><br /> <a href="docs/assets/demo.mp4">▶ watch the full-quality mp4</a> </p>

Install

Requires Bun ≥ 1.3.11, git, and at least one engine CLI on PATH. macOS, Linux, and Windows.

bun install -g @sma1lboy/rove

# or try it without installing
bunx @sma1lboy/rove

The global package installs both rove and kobe. rove is the canonical entry point, while kobe remains a compatibility alias. Supported legacy state is copied additively into ~/.rove; old files and worktrees stay put.

First run

ssh devbox        # optional
cd your-repo
rove

Press n, pick a repo, base branch, and engine, and prompt the embedded session. The worktree lands in ~/.rove/worktrees/<repo-key>/<task-slug>/. Press F1 anytime for the live keybinding reference; ctrl+q focuses the sidebar, and from there quits — sessions keep running in the background.

Full documentation: docs.kobe.sma1lboy.me — quick start, concepts, CLI, agent API, configuration, keybindings, themes, engines, plugins, troubleshooting.

If Rove saves you an afternoon, star the repo — it is the single strongest signal that tells other developers this is worth their time.

Beyond the three panes

Review with leverage. Open a file's diff, v a range, c a note, and s sends every unsent note across the whole task back to the engine as one prompt. The engine gets your words plus file and line numbers — it reads the worktree itself. Notes survive restarts, and sending doesn't switch tabs.

<p align="center"> <img src="docs/assets/diff-review.png" alt="Diff review — a range anchored across the agent's new lines and a note being written for the engine" /> </p>

Never babysit a rate limit. The footer carries each usage window the vendor reports (CLAUDE 5h 42% → 14:00 · 7d 12%). When Claude hits its subscription window, Rove schedules a resume and continues the task once the window resets. Or don't wait at all: ctrl+e continues the same conversation in a different engine — the next agent gets the previous transcript's path and picks up from there — and rove api send --tab new --vendor codex puts a second vendor on the same worktree, on the same files.

Run it from your phone. Below 70 columns the TUI becomes one panel at a time — task list and workspace alternate, the first row jumps you back where you were, quota shrinks to CLAUDE 42%. No setting, no separate app; it follows the terminal width.

<p align="center"> <img src="docs/assets/narrow-sidebar.png" alt="Narrow mode — the task list filling a phone-width SSH session" /> </p>

Work that runs without you. Routines are daemon-owned cron prompts: every firing creates a fresh task — worktree, branch, engine session — with the prompt as its first message. A --precheck command skips the run when nothing changed, so a nightly schedule doesn't burn a turn on an idle repo. An enabled routine keeps the daemon alive with no TUI attached.

<p align="center"> <img src="docs/assets/routines.png" alt="The Routines page — scheduled prompts with their cron expressions and next runs" /> </p>

And the rest, briefly:

  • Inbox (ctrl+a i) — what needs you, and where you were. F7 jumps to the oldest pending item across every project, even mid-typing inside a session.
  • Kanban (ctrl+a 1) — a local issue board where agents move their own cards (rove api issue-update --task), and you start a session straight off one.
  • GitHub issues (ctrl+a 3) — browse the repo's issues through gh and start a task on one; the body arrives as the first prompt, fenced and marked untrusted. Nothing is written back.
  • Field notes — rove api note files a resolved gotcha into the repo's durable store; every future worktree session on that repo starts with it in its system prompt.
  • Human in the loop — an agent can toast every attached UI (notify), ask you a question through the TUI and block on the answer (prompt), or open a split beside itself for a dev server (pane-open).
  • Attachments — drag an image or PDF onto a session and the path lands in its input; ctrl+v a screenshot and Rove saves it first. Rove only ever passes paths.
  • Themes and plugins — three bundled themes plus ten hosted (rove theme add <url>), and a manifest plugin system with panes, lifecycle events, chords, and an optional typed SDK.

Scripting it

rove api is the same daemon the TUI talks to, one JSON object per call — the surface a shell script or another AI agent drives:

rove api add --repo "$PWD" --prompt "Fix the flaky auth test."  # spawn a task
rove api read-output --task-id <id>                             # its own session history, paged
rove api send --prompt "succeeded: fixed, branch rove/auth"     # report home to whoever spawned you
rove api land --task-id <id>                                    # merge the winning branch

A task created from inside another Rove session records its dispatcher, so a bare send routes back to the exact tab that asked for the work — no ids to thread through. Install the companion skill and Claude Code drives this loop itself, with you at the gates:

rove skill install

Every verb, flag, and exit code: Agent API reference.

How it works

Task = git worktree + hosted engine session + branch
  1. The daemon owns tasks, worktrees, and state — the TUI, the web dashboard, and rove api are all clients of the same one.
  2. The PTY host is a separate long-lived process that owns the engine sessions. It outlives the TUI and the daemon, which is why disconnects and restarts never kill your work.
  3. The TUI just attaches: sidebar for tasks, workspace tabs for engines and shells (with splits and quick-fork), a files pane for diffs, review notes, and PR actions.

The same tasks are available from a browser — rove web serves a local dashboard on the daemon the TUI uses, so both surfaces stay in sync:

rove web            # http://localhost:45174

Rove vs desktop agent IDEs

Tools like Conductor and orca wrap parallel agents in an Electron desktop app. Rove makes the opposite bet: the terminal is the product.

| | Rove | desktop agent IDEs | |---|---|---| | Runs where your code lives | ✅ any box you can SSH into | local app reaching out remotely | | Survives disconnect | ✅ daemon + PTY host on the host machine | depends on the app staying open | | Install | bunx @sma1lboy/rove — zero-install trial | download a desktop app | | Agent-to-agent orchestration | ✅ rove api + companion skill | varies | | UI | terminal (TUI) + optional local web dashboard | Electron |

If your workflow is already SSH + terminal, Rove fits it instead of replacing it. If you want a desktop app, those tools are good — different bet.

If you write about developer tools, this README plus the landing page (which links right back here) should give you everything you need — the full manual is at docs.kobe.sma1lboy.me, and architecture details live in docs/ARCHITECTURE.md.

If it gets stuck

rove doctor            # read-only diagnosis: daemon, PTY host, engines, git
rove doctor --report   # write a bundle you can attach to a bug report
rove reset             # stop the runtimes; never touches your worktrees
rove config            # open Rove's config file in your editor

More in docs/TROUBLESHOOTING.md.

Develop

bun install
bun run dev:sandbox    # run against a throwaway home, not your real ~/.rove state
bun run test

Start with CONTRIBUTING.md and docs/ARCHITECTURE.md; shipped behavior lives in the changelog.

License

MIT © Jackson Chen

Related Skills

View on GitHub
GitHub Stars147
CategoryAI
Updated15h ago
Forks8

Languages

TypeScript

Trust signals

100/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

No cautions