Overview

What Brainlyy is, and the four commands that get you from nothing to a running instance.

A private brain for your machine, and an agent fleet for your company, reachable from any AI assistant that speaks MCP.

Your notes, your models, your codebase. No account, no API keys, and nothing leaves the machine unless you configure it to.

1. Install

uv tool install brainlyy

(or pipx install brainlyy / pip install brainlyy)

Not on PyPI yet. From a checkout:

uv build --wheel
uv tool install dist/brainlyy-0.1.0-py3-none-any.whl

Nothing else is required. Not a model runner, not a vector database, not a VPN, not a domain. See Concepts for why.

2. Set it up

brainlyy init

Asks at most three questions and defaults every one of them. Then, without asking: generates its own credentials, renders the knowledge and agent templates, registers itself with your MCP client, and runs its own property suite.

If that suite fails, init says so and does not report success. A setup command that claims "done" without checking is how people end up with a system that is confidently broken.

3. Look at it

brainlyy ui

Opens the dashboard: what tier you are at, what is missing, your notes with their provenance, and, if you have indexed one, your code graph. Binds loopback by default. See Security to reach it from a phone.

brainlyy doctor        # the same status, in the terminal
brainlyy search <q>    # query your notes without a browser

4. MCP server

init already registered it. Your assistant gains search_notes, ask_fast, ask_licensed and code_search. On the machine itself it also gains post_message, run_agent, agent_status and rate_answer.

Read tools bind your private network address. Write tools bind loopback and the process exits if configured otherwise. They hold different tokens, so the credential that travels your network cannot act. See Security.

More