Vicoa alternative: keep your agent workflow, add phone and Watch

The short answer

Vicoa ("Vibe Code Anywhere") is an AGPLv3 open-source agent orchestrator. A daemon or desktop app on your computer launches the CLIs of Claude Code, Codex, OpenCode and other agents, and the sessions sync to its iOS, Android, web and desktop clients. On top of that it has a task board, cron-style automations and team sharing.

BotBus differs in two places: where sessions come from, and who can read them. Vicoa manages sessions it starts; BotBus discovers the sessions you already open the usual way in a terminal, an IDE or the official desktop apps. BotBus also encrypts content end to end on your devices, and it has an Apple Watch app. If you need agents that start on a schedule, team seats and shareable session links, a self-hosted stack, or access from any browser, Vicoa is the better fit. If you want to keep launching claude and codex the way you do now, handle approvals from your phone or wrist, and care whether a relay can read your conversations, BotBus is worth a look.

Everything below about Vicoa comes from its own sources: the vicoa.ai homepage, pricing page and docs, its own Codex comparison page, the README of the vicoa-ai/vicoa GitHub repository, and its App Store listing. Checked on October 24, 2026.

What Vicoa is today

Architecture. Per the README, a local vicoa daemon "runs on your machine, next to your code" and spawns the agent CLIs. The daemon talks to Vicoa's backend (FastAPI with PostgreSQL) over WebSocket, and the web, desktop, iOS, Android and CLI clients all talk to that backend. The desktop app is Electron with a bundled daemon. Agents run with the CLIs and logins you already have, using your own subscription or API keys, and the site says there is no usage markup.

Starting sessions. Run vicoa in a terminal to start a session (Claude Code by default, --agent codex to switch), or run vicoa daemon to just connect the machine and start sessions from the phone or web dashboard by picking a machine, an agent and a directory. To pick a session back up you run vicoa --resume <session id>, and the docs say the IDs are listed in the Vicoa web app. As of the check date, the docs don't mention importing sessions that were started with plain claude or codex.

Agents. Built in: Claude Code, Codex, OpenCode, Gemini, Cursor, GitHub Copilot, Kimi, Hermes, Pi, Oh My Pi and Antigravity, plus 30+ more through the Agent Client Protocol (ACP), and other models through OpenRouter.

Features. Each agent can run in its own git worktree, with one list showing which sessions are running, blocked or done. There's a file explorer, a terminal next to the chat and per-file change review. From the phone you can chat, approve with one tap, switch permission modes, get push alerts and dictate. Tasks on the board turn into agent sessions. Automations are the standout: the site describes server-side cron schedules with frequency controls, a run-now button and run history, with agents reporting back when done. Sessions or whole projects can be shared by link, no account needed to view, with expiry and revocation; the docs warn that anyone with the link can read the full transcript and that Vicoa does not redact it.

Platforms. Desktop apps for Mac, Windows and Linux; the CLI installs via npm or pip. The iOS app needs iOS 15 or later and runs on iPhone, iPad and Apple Vision. There's an Android app and a web app. Neither the listing nor the docs mention Apple Watch.

Security. The FAQ says Vicoa sets up secure connections between your local environment and its infrastructure, and that your code stays on your machine; the relay carries the conversation, not the codebase. As of the check date, neither the docs nor the README describe end-to-end encryption or say whether the servers can read conversation content.

Pricing and license. Free: unlimited messages, one remote machine, ten automations. Pro: $12 per user per month, or $9 billed yearly, with unlimited machines and automations plus teams and shared projects. Enterprise is quoted separately and adds SSO, SCIM and on-premise deployment. The whole stack (CLI, daemon, backend, web, desktop and mobile clients) is AGPLv3, and the README documents self-hosting with Docker.

When Vicoa is the better choice

These are real strengths that BotBus doesn't offer today:

• Scheduled automations. Agents that start on a cron schedule and report back. BotBus has no scheduler; every turn is started by you or by the agent itself.
• Teams and sharing. Seats, shared projects and read-only session links. BotBus is a personal tool with no team features and no way to share a session with someone else.
• Any browser, plus a desktop IDE. A web app you can open on a borrowed laptop, and a desktop app with a task board. BotBus needs its app on the phone and runs as a background service on the computer, with no web client.
• Open source and self-hostable. You can audit, modify and run the whole thing yourself. BotBus always goes through our relay; the apps are closed source, and only the protocol, connectors and test fixtures are public.
• Older OS versions and more devices. iOS 15 and up, plus iPad and Apple Vision. BotBus needs iOS 26 on iPhone and macOS 26 on the Mac.
• More built-in agents. Gemini, Cursor, GitHub Copilot and Kimi are first-class in Vicoa. In BotBus they can only come in over ACP, and apart from OpenCode we haven't tested them one by one.

If any of those is a hard requirement, stay with Vicoa.

What BotBus does differently

No new launch command; existing sessions just show up

Vicoa's sessions start through its daemon. BotBus leaves your habits alone. For Codex it reads the local ~/.codex session data (read-only). For Claude Code you install hooks once from BotBus settings, and sessions you start in a terminal, an IDE or the Claude desktop app appear on your phone; on startup it also backfills the last seven days of sessions. Hermes, Pi, OpenClaw and DeepSeek Harness are built in too, and ACP agents such as OpenCode can be connected, see the connection docs.

On the phone you get tasks grouped by computer and project, each with a status: running, waiting for approval, waiting for a reply, done or failed. When an agent wants to run a command or edit a file, an approval card appears with Approve and Deny; multiple-choice questions are answered by tapping an option. You can follow up, interrupt or start a new task, and for Codex and Claude Code you can switch model and reasoning effort when you follow up. Whether each action is available depends on the connector, the task's state and whether the computer is online.

End-to-end encrypted content

When pairing, the phone scans a QR code on the computer, and the group key travels in that code, never through our servers. Tasks, conversations, commands, screenshots and push text are encrypted on the device before they leave it, and the relay running on Cloudflare only forwards and briefly stores ciphertext. The server can see device IDs, online status, update times and data sizes, and the dev preview that shows a local web page on your phone is not covered by end-to-end encryption. See the security notes, and for how other remote setups compare on who can see what, Claude Code Remote Control alternatives.

Apple Watch

The Watch app shows tasks, approves requests, answers questions and follows up by dictation; it syncs credentials from the iPhone and then talks to the relay directly. See managing AI coding agents from Apple Watch. Notifications fire only when an agent waits for approval or a reply, finishes or fails, and what goes to Apple's push service is ciphertext that the iPhone decrypts locally.

Remote access to the computer, and dev previews

Open "Control computer" from a chat to browse, edit and upload files, or open a terminal (your login shell on Mac and Linux, PowerShell on Windows). Terminal sessions live on the computer and survive the phone dropping off. The Mac version can also show the live screen and take taps and typing, for the system permission prompt or simulator an agent got stuck on. Local web pages an agent starts can be previewed on your phone, see preview a localhost dev server on iPhone. The screen and previews are Mac-only.

Several computers

One phone can pair with several computers at once, mixing Mac, Linux and Windows, with tasks listed per computer. The phone app is on iPhone (App Store) and Android.

Side by side

Using both, or switching

You can run both, with some care. Vicoa drives the claude and codex CLIs you already have, so the session data lands on your machine as usual, and BotBus should see those sessions too (Claude Code needs BotBus's hooks installed). We haven't specifically tested the combination, so treat it as something to verify. Answer each approval or question in one place only: if you approved in Vicoa, don't also tap it in BotBus. With push enabled in both you'll get duplicate alerts, so turn one off. If you rely on Vicoa's automations, one option is to keep Vicoa for scheduled runs and use BotBus for day-to-day sessions.

To switch to BotBus:

1. Install the computer app from the download section, install BotBus on your phone and scan the pairing QR code.
2. Install the Claude Code hooks from BotBus settings. Codex needs no setup.
3. Launch claude and codex directly from now on, without going through vicoa.
4. Anything you ran as a Vicoa automation needs another trigger, such as your OS's own scheduler. BotBus can't take that part over.
5. When you're done with Vicoa, stop its daemon or desktop app and revoke any session links that are still live.

Limits and trade-offs

• No scheduling, teams or sharing. Those are Vicoa's strengths and BotBus doesn't have them.
• Actions depend on the connector and task state. Some tasks are view-only. Approving, following up, starting and interrupting all need the computer online and an agent that supports the action. Pi has no approvals, for example, and sending images from the phone works only with Codex and Claude Code.
• The computer has to be on and online. Same for both: the agent runs on your machine and stops when it sleeps, see keeping your Mac awake for AI agents.
• Relay and app required. No self-hosting, no direct connection, no web client.
• Screen and previews are Mac-only. On Linux and Windows, "Control computer" means files and a terminal.
• Recent OS versions only. iOS 26 on iPhone, macOS 26 on the Mac.

FAQ

Is Vicoa open source?

Yes. As of October 24, 2026, its GitHub repository publishes the whole stack under AGPLv3, including the CLI, daemon, backend and every client, and the README explains self-hosting with Docker. The Enterprise plan adds on-premise deployment.

Can Vicoa's servers read my conversations?

Its docs say the connection between your machine and its infrastructure is secure and your code stays local, but as of the check date they don't describe end-to-end encryption or say whether the backend can read conversation content. If that matters, self-host it or ask the team. BotBus encrypts content end to end and the server sees only metadata, except for dev previews.

Does BotBus have scheduled tasks like Vicoa's automations?

No. Every BotBus turn is started by you, on the computer or the phone, or is part of an agent run already in progress. If you need agents to start on a schedule, Vicoa fits better. You could also call an agent's CLI from your OS scheduler, but how those sessions show up in BotBus depends on the agent, and we haven't tested it.

What other alternatives have you compared?

The series also covers the Happy Coder alternative, the Paseo alternative and the VibeTunnel alternative, one product each.