Getting StartedWorking with CerebroMCP ToolsSetup GuideConfigurationTroubleshooting
Documentation

Setup Guide

Download the installer and run it. Cerebro checks your system and handles prerequisites for you. One installer, done.

1Prerequisites

The setup wizard detects and installs these for you on first launch. You can also install them manually beforehand — the wizard will recognize them and skip to launch.

Claude Code CLIInstalled via the official claude.ai install script. Any Claude subscription tier works — Cerebro uses your existing session, no extra API cost. Codex (ChatGPT) or a local tool-use model work too.
Node.js LTSDownloaded from nodejs.org. Required by the Claude Code CLI.
Git (Windows only)Installed via winget. macOS and Linux ship with Git already.
Windows, macOS or Linux64-bit, 8 GB RAM minimum. The AppImage runs on most modern distros.

2What the setup wizard does

On first launch, Cerebro runs a prerequisite check before starting the main app. Here's exactly what it does, step by step.

Prerequisite wizard — consent-first

  1. System scan. Checks for Claude Code CLI, Node.js, and Git (Windows only). Each item shows a green check (found) or amber warning (missing).
  2. Per-item approval. For each missing item you get two options: [Install for me] or [Show manual instructions]. Nothing installs until you click a button.
  3. Official sources only. The wizard pulls exclusively from vendor-official locations — no Professor-hosted infrastructure involved.
  4. Launch. Once all prerequisites are confirmed, Cerebro opens. On subsequent launches the check runs in under a second.
  • Claude Code CLI — https://claude.ai/install.sh (macOS/Linux) or https://claude.ai/install.ps1 (Windows)
  • Node.js LTS — https://nodejs.org
  • Git (Windows) — winget install --id Git.Git
  • Git (macOS) — brew install git
Transparency by design. Every install action is visible, per-item, and requires explicit approval. Prefer to set things up yourself? Choose [Show manual instructions] for any item and skip the automated step.

3Install in four steps

From download to live

  1. Download the installer. Grab the latest build from your portal — Cerebro-Installer.exe on Windows, .dmg on macOS, Cerebro-Installer.AppImage on Linux. Every release is also on GitHub.
  2. Run it. Windows: double-click the .exe; SmartScreen may ask you to confirm — More info → Run anyway (why). Linux: chmod +x Cerebro-Installer.AppImage then run it — FUSE required (FUSE guide).
  3. Complete the setup wizard. It detects Claude Code CLI and Node.js; click [Install for me] for anything missing, or sign in to Claude Code manually if you prefer (below).
  4. Activate. Paste the passkey from your portal when prompted. Cerebro validates it once and you're live.
terminal
# Install Claude Code (if you prefer manual)
npm install -g @anthropic-ai/claude-code

# Sign in
claude login
That's it. Open the app and Cerebro is live. The platform auto-registers with Claude Code on first run.

4Optional: multi-machine memory

Want one shared memory across every device you use? Point Cerebro at a NAS-mounted folder. Recommended for power users.

  • Any NAS — Synology, QNAP, TrueNAS, even a Pi
  • An NFS or SMB share
  • A network connection on each device

Mount the share:

macOS / Linux
# Replace YOUR_NAS_IP and the share name with your own
sudo mount -t nfs YOUR_NAS_IP:/volume1/AI_MEMORY /mnt/nas
# Or persist via /etc/fstab:
# YOUR_NAS_IP:/volume1/AI_MEMORY /mnt/nas nfs defaults 0 0
Windows
:: Map a network drive (replace with your NAS IP / share)
net use Z: \\YOUR_NAS_IP\AI_MEMORY /persistent:yes

Point Cerebro at it. Open Cerebro → Settings → Storage and set the storage path to your mounted share (e.g. /mnt/nas/cerebro or Z:\cerebro). Repeat on every machine — they all read and write the same memory database. No sync needed.

5The creator's setup

Reference architecture for the home lab Cerebro is developed against. You don't need any of this — Cerebro runs great on a single laptop. This is what happens when you go all-in.

Windows 11 PC

Main dev machine
Development, Claude Code, daily driver

ASUS GX10

119 GB RAM · NVIDIA GB10 GPU · Ubuntu
Cerebro server — desktop app, backend, agents, operations engine

NVIDIA DGX Spark

128 GB RAM · GB10 GPU
Distributed embeddings, local LLMs, GPU-accelerated FAISS

16 TB Synology NAS

Centralized AI memory storage
All devices read and write here

Home server (Pi 5)

Raspberry Pi 5 · Ubuntu 24.04
Pi-hole DNS, Tailscale VPN, ambient services

A laptop is enough. The full home lab is for distributed embeddings and 24/7 agents. The same installer powers both.

6What Cerebro does not require

Common assumptions that are wrong — Cerebro is simpler than you think:

No DockerShips as a native installer with the runtime bundled — no containers, no Compose files
No Python or pipThe installer brings its own runtime — nothing for you to install
No API keysUses your Claude, Codex or local-model subscription directly
No cloud account100% local — your data never leaves your machine
No database serverBuilt-in SQLite + FAISS, zero config

Your operations platform is waiting.

An AI that remembers everything and runs your work — no API keys, no cloud, no per-token meter.

Get Cerebro