Installation

One-Liner

curl -fsSL https://the-brain.dev/install.sh | bash

The installer runs an interactive wizard by default — you choose which AI tools to harvest, which memory layers to enable, and whether to use MLX training or LLM backend.

For non-interactive installs (CI, automation):

curl -fsSL https://the-brain.dev/install.sh | bash -s -- --quick

The quick mode defaults to: cursor + claude harvesters, all 3 layers, LLM on, MLX off.

The installer auto-detects your OS:

macOS: Bun CLI + menu bar app + LaunchAgents
Linux: Bun CLI + remote mode with auth token

Reconfiguring Later

You can change your pipeline choices anytime:

the-brain setup                  # Re-run interactive wizard
the-brain setup --status         # See current configuration
the-brain setup --enable cursor,gemini --disable claude

Then restart the daemon:

the-brain daemon stop && the-brain daemon start

Prerequisites

Bun (curl -fsSL https://bun.sh/install | bash)
macOS with Apple Silicon (for MLX training — optional)
Python 3.11+ and uv (for MLX sidecar — optional)

Manual Install

git clone https://github.com/the-brain-dev/The-Brain.git
cd The-Brain
./install.sh

This handles everything: Bun checks, dependency installation, config generation, CLI linking, daemon setup, and LaunchAgents (macOS).

Initialize with a Project

the-brain init --project my-app --work-dir ~/projects/my-app

Each project gets isolated memory, wiki, and LoRA checkpoints.

Remote Mode (Linux Server)

# On server
the-brain init --remote
the-brain daemon start
# → API on 0.0.0.0:9420, MCP on 0.0.0.0:9422

# On Mac client
export THE_BRAIN_REMOTE_URL="http://<server-ip>:9420"
export THE_BRAIN_AUTH_TOKEN="mb_<token>"
the-brain agent

Verify

the-brain inspect --stats
# Memories: 0 | Graph nodes: 0 | Status: running