Skip to content
Docs0.9GitHubInstall

ReferenceFAQ

FAQ

Short answers to what people ask before and after installing Vendetta.

Short answers for anyone deciding whether Vendetta fits, or working out why it behaves as it does. Each links to the page with the detail.

No. Agents in any terminal are listed and can be messaged, stopped, and scheduled, and Claude Code, omp, and OpenCode take prompts natively.

tmux adds pane navigation, live terminal previews, and managed agentsManaged agent. An agent Vendetta started, in its project's Agents tab. Agents you start by hand are listed too and stay where you started them.. Codex, Gemini CLI, and Aider need tmux or Kitty for remote prompts. See Without a host.

Nothing, until you turn something on.

You turn on What it sees
Telegram or Discord What those chats show. Bot chats are not end-to-end encrypted.
A transcription provider Your voice notes
ntfy or push Alert text

The web client and the iPhone app connect to your own machine. The app port listens on loopback until you publish it, and every connection needs a paired device. Security lists every connection and what turns it on.

Claude Code 2.1.296 or newer, Codex, omp, OpenCode, Gemini CLI, and Aider 0.86.2. What each one gets differs: questions arrive as buttons from Claude Code, omp, and OpenCode, and Gemini CLI and Aider are listed but not started by Vendetta. Providers has the table.

Vendetta brings no agent of its own. You bring the providers and their logins.

For Claude Code, yes, in the iPhone app or the web client. The chat shows a card with the command or edit and why Claude asks, with Allow once and Deny.

Other providers’ dialogs, and a Claude subagentSubagent. An agent a Claude Code session starts inside itself. Its permission dialogs are answered at the terminal.’s, you approve at the terminal, which the chooser can reach over SSH. Telegram, Discord, agents, the Master, and schedules never answer an approval.

Do agents that Vendetta starts ask before running tools?

Section titled “Do agents that Vendetta starts ask before running tools?”

On your computer, yes: an agent Vendetta starts asks before it runs a command. In a container project it doesn’t ask, and the container is the boundary.

Agents you start by hand keep their own settings. To change what Vendetta’s agents start with, see Approvals.

No. Agents run as you, with your permissions, and so do plugins. A project machine runs a project’s agents in a Docker or Podman container on a private clone, and there the container is the boundary.

Yes, on TestFlight. The iPhone app says how to get it, how to build it yourself on a Mac, and what push notifications need. The web client works in any phone’s browser and needs no install.

Yes, on macOS 13 or newer, installed the same way as on Linux. See Install.

Do the people I add need their own provider account?

Section titled “Do the people I add need their own provider account?”

Not unless you say so. A person’s agents use your Claude login through the Claude proxy by default, and count against your plan’s limits. A consumer Claude subscription is for one person, so read People first.

To have a person sign in with their own account inside their machines, add them with --claude-login own. Each person connects their own GitHub from Settings → Connections.

What happens if I message an agent while it’s working?

Section titled “What happens if I message an agent while it’s working?”

It gets your message as a normal prompt, the same as typing it in its terminal. Vendetta doesn’t interrupt the turn, and a refused message is never replayed later.

In the apps you can also queue a message for after the agent’s turn.

No. The Master is optional, and nothing else depends on it.

Its integration isn’t loaded. Install it, then restart the agent or trust its hooks. See Provider integrations.

What does Vendetta store about my conversations?

Section titled “What does Vendetta store about my conversations?”

Its history journals hold states and metadata, never prompts, messages, or tool arguments.

Chat history in the apps is read from each agent’s own transcript. The app plugin holds messages in memory only, and keeps private copies of the photos, videos, and audio a chat shows.

A schedule failed. Will it run again by itself?

Section titled “A schedule failed. Will it run again by itself?”

No. A failed or uncertain run stops the schedule until you check it. The same rule holds everywhere: when Vendetta can’t tell whether a prompt, a launch, or a deletion happened, it says so and never repeats it. See A nightly job with run history.

Yes. A plugin is a separate program on the worker’s socket, in any language. It can react to agents, serve methods, and add pages and buttons to the apps. See Write a plugin.

AGPL-3.0, with a commercial licence available from the author. See Licence.