Chatbox Community Edition

What Chatbox does

Chatbox is a desktop chat client: you add API keys for the model services you use, or point it at a local model server, and talk to all of them from one window. Beyond plain chat it handles files and images, keeps a local knowledge base of your documents on the desktop, searches the web, connects to MCP servers and has a Work Mode in which an agent can run code, read and write files and hand back downloadable results.

Community Edition and the official app

The GitHub repository is the Chatbox Community Edition under GPL-3.0. The project was closed source for about two years before it was reopened, and the README says code is synced between this repository and a private “pro” repository in both directions. The repository holds a complete, buildable Electron and React application, with commands to run and package it yourself and a web build.

The installers on chatboxai.app, and those linked from the GitHub release notes, are the company’s official builds. The project does not document what, if anything, differs between those builds and the Community Edition source. If you want to be certain you run only the GPL code, build it from the repository. The company also sells Chatbox AI, a hosted model subscription; the client code for it is in the repository, and some conveniences use it, such as the built-in search option and a hosted document parser for paying users.

Models and providers

The source defines more than 30 providers, among them OpenAI, Azure OpenAI, Anthropic, Google Gemini, AWS Bedrock, DeepSeek, Mistral, Groq, xAI, OpenRouter and Perplexity, plus custom OpenAI-compatible endpoints. For local models it connects to Ollama (default address http://127.0.0.1:11434) or LM Studio; Chatbox has no inference engine of its own. Prompts go directly to whichever provider you select, so using a hosted API sends your conversation to that company.

Privacy and data

Conversations stay on your device: on the desktop, settings are kept in a file and chats in IndexedDB, and knowledge base vectors are stored locally. Analytics are on by default. The repository’s design notes describe Google Analytics with a persistent install ID, Plausible event tracking and Sentry error reporting, all controlled by the reporting and tracking checkbox under Settings > Advanced. Unticking it stops them.

Running Chatbox

Official builds support Windows 10 (x64), macOS 11 or later on Intel and Apple Silicon, and Linux x64 via AppImage or Ubuntu 20.04+. To build the Community Edition you need Node.js 22 and pnpm. Chatbox is a client, not a server, so nothing needs hosting. The optional team-sharing folder includes a Docker and Caddy setup that lets a team share one OpenAI key without handing it out.

Hosted version

The company runs a web version of the app and the Chatbox AI subscription for hosted models, with a free daily quota for new users. Neither is needed if you bring your own keys or use local models.

Limitations

The open source edition and the official binaries come from different repositories, and the difference is not documented. Telemetry is opt-out rather than opt-in. Local models need a separate runtime. Most of the design documentation in the repository is in Chinese.

Who it suits

Chatbox fits individuals who want one polished desktop client for several paid APIs and local models, with files, search and agent tools included. Readers who mainly want a local model runner should look at Jan, and teams that need shared accounts on a server at LibreChat.