What Actually Happens When You Run npx nexo-brain
This is the real product flow: not just “MCP setup”, but runtime installation, shared-brain configuration, client choice, automation choice, and recommended model profiles.
$ npx nexo-brain
Language, path, operator identity
The installer asks for language, where the runtime should live, and the operator identity NEXO should adopt. Shared brain is always on.
Environment scan
NEXO checks what is already installed on the machine and looks for supported clients such as Claude Code, Codex, and Claude Desktop.
Shared-brain clients
You decide which interactive clients to connect. Claude Code remains recommended, Codex is supported, and Claude Desktop can share the same brain through MCP.
Terminal default
NEXO asks which client nexo chat should open by default, so you do not have to remember separate launch commands later.
Automation backend
You choose whether background automation is enabled, and which backend should run it. Claude Code is still the primary recommended automation path.
Model profiles
Recommended profiles are applied at install time: Claude Code → Opus 4.6 with 1M context, Codex → GPT-5.4 with xhigh reasoning.
Recommended path
Install the runtime, use Claude Code as the default client, keep automation on Claude Code, and let Codex or Claude Desktop attach to the same shared brain when needed.
After install
The runtime ends with the three commands that matter most ready to go: nexo chat, nexo update, and nexo doctor.
If you add clients later
You can always resync them without reinstalling: nexo clients sync.