The primary story is deliberately simple: run one command, let NEXO set up the local runtime, wire the shared brain for the clients it finds, and verify the runtime immediately.
npx nexo-brain installs the runtime, downloads the embedding model, configures the shared-brain MCP layer, and asks which clients and automation backend you want active.
nexo doctor gives the fastest proof that the runtime is healthy, the managed schedules are coherent, and the local brain is actually there.
nexo chat launches the configured terminal client instead of assuming one forever. That matters when you use both Claude Code and Codex.
NEXO sets up the operational layer around memory too: doctor, schedules, local databases, shared-brain client sync, and the runtime CLI.
The install path keeps vectors, memory stores, logs, and workflow state on your machine. There is no hosted control plane hidden behind the installer.
For new people this is the easiest proof point: one command, a visible health check, and a client they can launch immediately.
After install, the next thing people want to see is whether the shared brain really survives across client surfaces.