The landing page tells the story. This page is the reference: every capability in the shipping app, organized by what it helps you do. Where something is Mac-only or still in TestFlight, it says so right there.
Connect a server and iVibecode finds the coding-agent sessions already running in tmux on it — no matter which terminal started them — and keeps the list live.
Each session shows where it stands at a glance: Needs you (waiting on an answer), Working (busy), or Done. You scan the list, not ten terminals.
Sessions get readable titles recovered from the agent's own conversation — what it's actually working on — instead of raw identifiers. When nothing useful can be recovered, you get a clean "Untitled tab" label, never a raw machine identifier.
Each person normally sees only the sessions started by the account they connect as. On a shared server, an administrator can turn on "Show everyone's sessions" to see teammates' sessions too — useful for covering for someone or keeping an eye on a fleet you operate together.
Discovery scans every tmux socket on the server, not just the default one — so sessions started by launchers and wrappers are found alongside the ones you started by hand.
Session cards carry enough context — status, title, recency — to decide "does this need me right now?" without opening anything.
A session that someone else is currently attached to says so, which avoids two people typing into the same agent at once.
Each server entry shows its own connection status, so when one box drops you can see exactly which one — the rest keep working.
Finished, detached conversations fold into one collapsed group, and rows for sessions that have genuinely died are pruned automatically — the list stays honest.
Open a session and, instead of raw terminal scrollback, you can read it the way it actually happened: your prompts, the agent's replies, its progress — clean and scannable.
Prompts, replies, and images appear as a clean conversation. The spinners, redraws, and control characters that make terminals unreadable are filtered out.
When the agent is working through a task list, a progress strip shows how far it's gotten — without reading a word of the log.
A session that's waiting on you says what it's waiting for — a summary of the question, right where you'd answer it.
When an agent offers numbered options ("1. merge, 2. hold off…"), iVibecode summarizes what's being asked and recommends an answer you can apply with one tap — no counting lines in a terminal.
Reading view is a toggle on each session — and starting with 0.5.0 it opens by default for new sessions. The raw terminal is always one tap away.
The reading view distinguishes "this conversation finished" from "this server is unreachable" — two very different situations that raw terminals blur together.
Long code lines scroll sideways inside their block instead of wrapping into mush, and leftover terminal artifacts are cleaned out of the transcript.
Open a session that isn't running an AI agent and the app falls back to the normal terminal view instead of pretending there's a conversation to show.
Watching is half the point. The other half: when an agent needs you, you can handle it in seconds — from the couch, the train, or another machine.
Type a reply and it lands in the running session as if you were at the keyboard — answer a question, change direction, approve the next step.
Speak your reply and iVibecode transcribes it into the session, powered by the same engine behind Expertly Hear.
Send a message to a busy Claude Code session and Claude's own type-ahead queues it, then delivers it when the agent is ready — iVibecode shows you the queue so you know your message is waiting.
Sessions that need you are summarized in one place, with the agent's question and — when it offered numbered options — a recommended answer you can send with one tap.
Start a reply, get distracted, open another session — your unsent text is still there when you come back.
Paste a screenshot into a session and the image is transferred to the server with its path typed in — the agent can look at what you're seeing.
Laptops sleep, trains hit tunnels, servers reboot, agents exit. iVibecode treats all of that as normal weather, not a crisis.
When the connection drops — even right as a session is starting up — the app keeps retrying and brings it back on its own. Stale "reconnecting…" banners clear the moment the link is healthy again.
A conversation whose session has ended isn't gone: iVibecode lists recoverable conversations and can resume one into a fresh session, picking up where it left off.
Claude, Codex, Gemini, and Qwen sessions each reconnect to their own real conversation after a drop — through the right tool for that agent — instead of coming back as a blank new session.
Attaching to a session where the agent already exited recovers the recent shell history above the live prompt, so you can see what it did last instead of a blank pane.
Sessions that have genuinely ended disappear from workspaces and lists automatically — you're never clicking rows that only error.
Your servers and workspaces are backed up on a rolling basis, and a damaged saved-server entry costs you that one entry — not your whole server list.
Group, find, and lay out your sessions the way you think about the work — and have it all still be there tomorrow.
Create named workspaces for a feature, a client, a fire drill. Drag sessions between them. Quit the app, reboot the server — the workspaces are still there.
Pin the sessions you keep coming back to, and let the recents list catch everything else — ordered by the last time a human actually did something in them.
Open a workspace with one session and it fills the pane; a workspace with several shows them as a clean list of large rows, each one tap from its full-size terminal.
Open sessions in their own windows, arrange them across displays, and relaunch — the windows are restored where you left them.
Search open sessions, past conversations, or both — with live match counts across sessions. Jump straight from a search hit into the session, or open the result in its own window.
Collapse an actively-working session down to its title bar to reclaim space; typing into a compressed window expands it again automatically.
iVibecode's whole security story is that it doesn't add anything: no server agent, no new open port, no relay in the middle. Your repositories stay on your servers — iVibecode connects over SSH and never copies your repositories. The AI status summaries send only the last screenful of session text to Expertly's AI service to write the gist.
Add every box your agents run on — a Mac in the office, a cloud virtual machine, the big shared development server — and watch them all in one list. Mark one as the default for the sidebar's new-session button.
Host aliases from your existing ~/.ssh/config auto-fill when adding a
server. iVibecode reads your config; it never edits it.
No key on this device yet? Give the app your server password once — it installs a device-specific key over that connection and does not keep the password.
Each device gets its own SSH key, kept in a private, permission-locked store on that device. Lose a device, revoke one key — not your whole setup.
Servers are verified by their host key, so a machine that suddenly "changes identity" is flagged instead of silently trusted.
Because it's plain SSH, it works over whatever path already reaches your servers — including private networks like Tailscale or a VPN. Nothing new is exposed to the internet.
Sometimes the fastest keyboard is none at all.
Hold to talk, and your words are transcribed into the running session — powered by the Expertly Hear engine. Long answers stop being a chore.
Say what you want done and iVibecode routes the request to the right session — or kicks off the work — without you hunting through the list first.
Most remote tools understand exactly one vendor's agent. iVibecode is built for the mixed fleet that real teams actually run.
Each agent type is recognized and labeled as what it is — a Codex session is never mislabeled as Claude. Claude Code, Codex, and Gemini each have their own dedicated parser; Qwen Code is a Gemini fork, so Qwen sessions ride Gemini's.
When a session drops, each agent is resumed through its own tool into its own conversation — Claude, Codex, Gemini, and Qwen alike. One honest note: Gemini and Qwen resume only on servers where their launcher tooling is installed.
A Claude refactor, a Codex bug hunt, and a Qwen docs pass sit side by side in the same list, same statuses, same reading view. You stop context-switching between vendor-shaped windows.
Start a fresh session on any server and pick which agent runs it — right from the app.
iVibecode updates frequently — see the changelog for the honest week-by-week history.
When a new version ships, the Mac app downloads and installs it itself (built on Sparkle, the standard Mac updater). There's also a manual "Check for Updates" if you like pressing buttons.
Every update tells you what it fixes in human terms — what was going wrong for you and what's different now — not commit-log shorthand.
Settings shows the exact build date and version, so "am I running the fix?" takes one glance to answer.
When something breaks, you can send us a privacy-scrubbed diagnostics report — no code, no credentials, no session content — so we can fix it fast. Automatic error reporting, with an off switch in Settings, is being added in the current release.
A signed, notarized Mac installer with automatic updates. Requires macOS 15 or newer. Sign in with your Expertly account and you're in — no credit card required. Don't have an account yet? Join the waitlist and we'll get you set up.
The iPhone and iPad app is in TestFlight early access today, with the App Store listing pending. Join the waitlist and we'll send you an invite.
A browser version that needs no Mac at all is in the works. Not available yet — when it is, the changelog will say so.
The watching side needs nothing installed: if your agents run in tmux and you can SSH in, iVibecode can see them.
A touch-friendly bar above the keyboard supplies the keys terminals need — Esc, arrows, Ctrl — so a phone can genuinely drive a terminal.
Drag a file onto a session and it's uploaded to the server with its remote path typed in — no manual copying.
Pick the look you want for the terminal.
A short guided tour explains the core ideas on first launch — and you can re-run it any time: Help → Guided Tour on Mac, or the About screen on iPhone.
Under the reading view there's still a real, faithfully rendered terminal — resized to fit your screen — for the moments you want the raw thing.
Keyboard shortcuts are shown in tooltips and menus, so you learn them by using the app.
| Shortcut | What it does |
|---|---|
| ⌘N | Start a new session — on the server you're viewing, or your most recent one |
| ⌘F | Find across sessions and past conversations |
| ⌘D | Do Something For Me — describe a task, iVibecode dispatches it |
| ⇧⌘C | Compress the current working session to its title bar |
| ⌘-click | Open any session row in its own window |
Free during early access. Connect a server and you'll be watching your agents in minutes.
Download for Mac