MengBotDocs
Getting started

Installation

Install MengBot on macOS, Windows, or Ubuntu.

Download a released build

Choose your computer below. Each link always downloads the latest published build.

PlatformRecommended packageNotes
macOS Apple siliconDownload .dmgSigned and notarized. Drag to Applications.
macOS IntelDownload .dmgSigned and notarized.
Windows x64Download installerPer-user installer; Windows signing is not yet available.
Ubuntu 24.04 x64Download .deb · AppImageThe .deb is recommended; the AppImage is portable.

Windows SmartScreen

The Windows installer is not code-signed yet. Windows may show “Unknown publisher.” Verify that you downloaded it from the official MengBot releases repository before choosing More info → Run anyway.

Install an agent CLI

MengBot needs at least one supported agent engine installed and signed in. Claude and Codex are the most common starting points.

  1. Install the CLI from its official provider.
  2. Run it once in a terminal and complete sign-in.
  3. Restart MengBot.
  4. Open the model picker. Available engines appear normally; unavailable engines explain what is missing.

See Agent engines for the complete list and explicit path overrides.

Using MengBot at work

If your company uses OpenMeng Admin, you can sign in with your organization to use the AI models it provides, alongside your own. It is optional. See Use MengBot at work.

Build from source

Requirements: Node.js 24 or newer, pnpm 10.33.0, and at least one authenticated agent CLI.

git clone https://github.com/EternalNight996/mengbot.git
cd MengBot
corepack enable
pnpm install --frozen-lockfile

Run the harness, web UI, and Electron shell in separate terminals:

pnpm dev:server
pnpm dev
pnpm dev:desktop

Where data lives

MengBot stores its application data under ~/.mengbot. Electron window and browser data use the normal per-platform application data directory.

On this page