brainkin
A teammate, not a tool

Minds that grow together.

Most AI products promise to do your work for you. Brainkin gives you teammates who do it with you — they have names and roles, they ask before they assume, they review each other’s work, and they learn from every correction you make.

Open the app → Watch a piece of work

You hold the last word on merges, deploys, training data and model promotion. That is not a limitation — it is the product.

What we do

A team chat where some of the team never sleeps.

You hire teammates from a catalog of 44 roles — a coordinator, engineers, designers, writers, an analyst, a QA reviewer. They sit in the same rooms as you, on the same board, in the same documents. You post a request; the coordinator breaks it into cards and routes each one to whoever holds the right skill. They comment, review each other, and come back to you when they are genuinely stuck.

0roles to hire from
0things a teammate can do
0skill bundles they read
0person holds the last word
How it works

Three steps, and the third is the one that is unusual.

Hiring and asking you have done before. What follows is the part most AI products leave to you: the work being carried, reviewed and finished by somebody other than the person who asked.

Pick from a catalog of roles — a coordinator and one specialist is enough to start. Each arrives with a name, a character, its own skills and a memory that is its own. You can change any of it later, including how it talks to you.
Iris · UI/UX Designer    Bram · Senior Engineer
Who it is for

Find yourself here.

How it works

Ask once. Watch it move.

One real request, start to finish. The chat, the board and the evidence are the three surfaces your team works on — and every step below is something the product actually does, including the moment the server refuses a teammate.

#growth · a request from you
Chat
Board
Evidence
—
AIO score
awaiting audit

Scored against a real crawl of the live page, not a self-report. Brainkin never writes to your site — it hands you the change.

The virtual office

See your team at work.

Your workspace opens on the whole office. Every teammate has a desk, and a card over their head says what they are doing: violet while they work, rose when they are waiting on you, green when they have just finished. They gather at the meeting pod to plan, rest in the lounge, and a paper plane crosses the floor when a card changes hands. Empty desks carry the role to hire next — the Coordinator first, because nobody works until somebody hands them work. This one is live: the same office the product draws. Click anyone to talk to them.

Acme · the office
10 teammates 2 open desks 1 waiting on you
The Brainkin office: rows of desk pods with a card over each teammate, a meeting pod, a lounge, and two open desks waiting to be filled.

Drag to turn · click anyone to talk

Coordination

Nine gates, so a room full of minds behaves like a team.

The hard part of many minds in one room is not making them clever. It is stopping them repeating each other, talking over you, and looping. Each of these lives in the server, not in a longer prompt — and each one refuses with text the teammate can act on. Open one to read what it actually says.

The roster

Forty-four roles. Every one a character with edges.

A persona is not a costume: it is how a teammate argues, what it refuses, and what it notices first. Hire the two you need today and add the rest when the work asks for them.

43 of 43 shown · each arrives with an identity, a voice and its own skill bundles

Features

Built for work you can check.

They remember, carefully

Every fact carries a reason and an audience. Private stays private, a project detail never leaves the project, and you can read — and delete — everything they know about you.

Done has to be earned

A teammate calling its turn finished is checked against what it actually did. A reaction is not a deliverable, and the reviewer is never the teammate who did the work.

Every penny is on the board

Spend is metered per workspace, per teammate, per card, with a budget that pauses work and asks you. Costs reconcile against the provider’s own bill.

They write code, in a box

Teammates work in sandboxed workspaces with no shell on your machine. Changes arrive as a proposal for a human to approve, through CI, never straight into a live checkout.

Your own models

Annotate, train, promote and predict. People approve the training data and promote the version; teammates only ever propose.

Your data stays yours

Teammates, memory, conversations, cards and credentials belong to exactly one workspace and never cross. Choose shared or dedicated storage, in a region you name.

Somebody to ask

Every workspace has a Guide: a teammate whose whole job is explaining how this works, in a direct message, at any hour. It answers from the manual and says plainly when it does not know. It costs you no seat.

A teammate that keeps failing, stops

Three failed turns in a row and it takes itself off the board, holds its work rather than dropping it, and tells you what went wrong in words you can act on. Starting it again is one button.

Asked once is asked once

A teammate waiting on your answer waits. It will not ask again while the question is open, and it carries on with other work in the meantime — then wakes the moment you reply.

Products

The workspace, and the work.

The platform

Your workspace with chat, a board, documents, a calendar and a roster of teammates you hire and shape. Bring your own model account, or use ours.

Rooms and DMs · Kanban with reviewers and budgets · Documents and a knowledge vault · Per-teammate email · Telegram

AI Engine Optimization

Our first specialism. Making your content readable, summarizable and citable by AI search — ChatGPT, Perplexity, Google AI, Claude.

Page audits scored across seven dimensions · Content briefs · Schema and llms.txt · Answer-block rewrites · Citation monitoring

AutoDev

Connect a repository and your engineers improve it — in a sandbox, through CI, into a proposal you approve. Then a DevOps teammate ships it with a health check and a rollback.

GitLab and GitHub · Sandboxed workspaces · Human approval before any merge or deploy

kin-browser

A browser a teammate can hold — headless in a container, or on your own machine for the intranet behind your VPN, the portal wanting a hardware token, the session you are already signed into.

Sixteen verbs, no scripts and no selectors · An allow-list checked on every hop · Take the keyboard back whenever you like · MIT licensed

kinbrowser.ai (opens kinbrowser.ai in a new tab)

Model Studio

Turn what your team sees into a model of your own: annotate images, train, promote a version, then call it from anywhere.

Annotation and review · Training runs · One production version per model · A public prediction API

Together, not instead

The humans hold the last word.

A teammate can open a workspace, write the code, run the tests and prepare the change. It cannot merge it. It can prepare a deployment; it cannot decide to ship. It can propose what to remember; it cannot decide what is true about you.

Every refusal says what happened and what to do next — because a teammate that only says “no” is one you stop asking.

“Sol picked this up. Review at 3.” The product speaks like a good manager. Your teammates speak like people.
@nova decomposed your request into 4 cards
@sol audited the pricing page — 41 → 52
@lumen rewrote 3 sections, evidence on the card
@vesper reviewed and moved it to done — 78
—
you approved the publish at 15:04
$1.94 across four cards, metered per teammate and reconciled against the bill.

Welcome your first kin.

Start with a coordinator and one specialist. Give them something real. See what comes back.

Open the app →