One desktop app for every coding agent you pay for.
Codex, Claude Code, Cursor, Grok Build, OpenCode, Antigravity, Droid, Kilo, Pi — nine CLIs, one dark-first Electron workspace on Linux, Windows, and macOS. Chat, terminal, preview, git, pull requests, stats, themes. Bring your own subscriptions. Ronin sells you nothing and phones home to no one.
If the CLI works in a terminal, Ronin can drive it.
The good stuff
🔄 Switch providers mid-thread
Start with Codex, hit a wall, hand the thread to Claude. Same history, same checkpoints, same working directory.
An agent that's been here before resumes its own session. A newcomer gets a brief built from the thread. Switch back later and it picks up where it left off — no copy-paste, ever. (docs)
🥊 Get a second opinion
Same prompt, two (or more) models, each in its own worktree. Press Compare, pick who should answer, and read them as ordinary threads with chips to jump between. Keep the one you like; delete the rest. (docs)
👥 A team, not just a model
Claude leads, Codex implements, Grok reviews — or any roster you name. You define the roles; the lead hands work around in one shared worktree until the task is done. Gate a role if you want to see it coming. Not the default; you build the team. (docs)
💬 Ask on the side
Select the passage you're stuck on. A chip appears — Ask on the side opens a fresh thread on exactly that text, same project, same checkout. The main thread never learns you asked. (docs)
📌 Capture as you read
Agents leave work behind in the transcript — "this helper could go," "these two want merging." Select it and press Capture. Ronin files a draft thread, ready on every device, and you stay where you were reading. (docs)
📋 A board, not just a list
Six lanes — Draft, Up Next, Working, Needs You, Snoozed, Done. Drag to settle, snooze, or pick the next thing. The sidebar and the board never disagree. (docs)
▶️ Replay a turn
Come back to a twenty-minute turn and press replay. The prompt, each tool call, and the reply play back in order, pauses compressed so you see the rhythm without sitting through the dead air. (docs)
📊 Stats that count everything
Reads each provider's own session transcripts — every token, including turns you ran in a bare terminal, on every connected machine. Real costs (Grok's exact figure, not an estimate), cache savings, per-model breakdowns, an hourly chart, and a sidebar meter that shows when your window resets, to the minute. (docs)
⏳ Hit a limit? Ronin waits
A spent five-hour window isn't a failure you can act on. Ronin parks the message, counts down above the composer, and sends it the moment the window resets. Cancel it, fire it early, or walk away. (docs)
🖼 Previews, inline
When an agent writes HTML — a coverage report, a chart, a mockup — Ronin renders it in the transcript. Sandboxed. Relative assets load. No "open this link." (docs)
🔀 Git as a first-class citizen
- Checkpoints after every turn — diff or restore anything an agent did
- Commit, branch, discard, worktrees in the topbar
- Pull requests across every connected server — review, react, edit in place
🎨 Themes that go deep
Tsukimi, Aizome, Urushi, the pure-black OLED Void family — plus Open VSX theme search. Pull almost any VS Code theme straight in. Import your own. Light and dark stay intact.
🧰 Skills once, every agent
Drop a SKILL.md in ~/.ronin/skills and every provider can use it. Slash commands — /clear, /compact, /model, /review, /fork — work the same way, next to each CLI's native ones. (skills · commands)
⏰ Work that runs without you
Save a prompt, pick a schedule, walk away. Each run is a real thread you can read afterwards, defaulting to its own worktree so unattended edits never land in your checkout. (docs)
📱 Remote from anywhere
npx t3 pair --tailscale
Scan the QR. Your phone is driving the same threads, over your tailnet, on hardware you own. LAN and SSH work too. (docs)
And the rest
- Since you last looked — what finished, what's still running, what's waiting on you (docs)
- Needs you — a sidebar queue, and a dock badge, for anything that stopped until you answer
- Agent notifications — a system ping when a turn finishes, fails, or waits for approval
- Native on Windows — no WSL; the same CLIs you run in PowerShell
- Live browser preview — tabs, navigation, screenshots, recordings
- Permission modes — decide how much rope each agent gets
- Transcript export — Markdown or JSON
- Editor hand-off — opens any file in the editor you actually use
Gallery
Run it
[!IMPORTANT]
Install and authenticate at least one provider first:
Desktop
Download the latest build from Releases. The app updates itself.
Or build from source:
# Node 22.16+ / 23.11+ / 24.10+
curl -fsSL https://vite.plus | bash # install vp (macOS/Linux)
vp i
vp run dev # server + web
# or
vp run dev:desktop # Electron shell
vp run build:desktop
vp run dist:desktop:linux # Linux AppImage
vp run dist:desktop:win # Windows installer (NSIS)
vp run dist:desktop:dmg:arm64 # macOS
Artifacts ship as Ronin-${version}-${arch}.${ext}. On Windows, Ronin runs native CLIs — no WSL. (install docs)
Documentation
Full tree: docs/.
Built on T3 Code.
Early. Expect bugs. Prefer small fixes over grand PRs unless we ask. Read CONTRIBUTING.md first.
