Frequently Asked Questions

What is Jean2?

Jean2 is an open-source AI agent platform. Connect your own LLM, define tools in any language, and run from a single binary. Desktop, browser, and mobile clients are available.

Is Jean2 free?

Yes — the core is open source (Apache 2.0) and free to use. You only pay for your LLM API calls.

What LLM providers does Jean2 support?

LLM providers include OpenAI, Anthropic, Google, OpenRouter, MiniMax, Zhipu, and Codex (ChatGPT subscription). You can connect multiple providers and switch models at any time — even between messages in the same conversation.

How is Jean2 different from Claude Code or Cursor?

Jean2 is not locked to one vendor or LLM. You can write tools in any language. It has MCP support, subagent orchestration, and runs entirely on your machine with no hidden system prompts.

Does Jean2 require an internet connection?

The server runs locally. You need internet for LLM API calls, but it works offline for non-LLM tasks. Remote access is available via Tailscale or VPN.

Can tools be written in any programming language?

Yes. Tools are scripts that speak JSON on stdin/stdout — Rust, Python, Bash, Go, TypeScript, anything. Drop them in and the agent picks them up.

What are subagents?

Subagents are isolated child sessions for complex tasks. Each has its own tools, model, and workspace context. Interrupts cascade automatically.

How do I access Jean2 from another device?

A REST API + WebSocket is built in. You can connect over local network, via Tailscale, or through a reverse proxy with TLS. All clients connect to the same server.

What platforms does Jean2 support?

Server runs on macOS, Linux, and Windows. Desktop clients are available for macOS and Windows. Browser access works on any platform. Mobile support is coming soon.

Does Jean2 inject hidden prompts?

No — this is a core design principle. The system prompt is entirely yours. Jean2 only injects available subagent info when needed.

Does Jean2 support images, files, and video in conversations?

Yes. If your model supports reading images, files, or video, drop them directly into the conversation. If you switch to a model that doesn't support a particular input type, Jean2 automatically replaces the media with a text reference to the file on your server — you never lose context. You can also drop in any text-based file and ask questions about it, which works with every model.

How does Jean2 handle security and privacy?

Jean2 never collects or sends any data. The only external communication is with your LLM provider. Communication with your server is encrypted with TLS. You authenticate with an API token. Each tool can define its own security level — the agent will prompt you to approve or deny access to anything outside the workspace or to sensitive files like .env.