Enter your email to signup for updates and your download will start automatically.
BrainMocker is a desktop app that runs a team of AI agents for you. The built-in Chief of Staff agent breaks down your goals into tasks, assigns them to specialist agents on a Kanban board, and only checks in when your rules say so.
You stay strategic. The Chief of Staff handles the choreography between agents, the file system, and the board.
Point at any folder on disk — empty, half-built, legacy. State the outcome you want in plain English.
Tasks, owners, priorities — proposed in a Kanban board you can edit before a single token gets spent.
| Task | Agent | Pri |
|---|---|---|
| Scaffold React + TS | fullstack | high |
| Design tokens | designer-ux | high |
| Core todo state | fullstack | med |
| Accessibility audit | qa | low |
Each task gets its own AI agent that can read your files and run commands on your computer, just like you would. You watch progress live on the card.
The Chief of Staff is built in. Worker agents are simple text files in your workspace — give them a name, describe their job, and pick what they're allowed to do.
Reads what you want done, drafts a plan, and hands tasks to the right specialists. Checks every agent question against the rules you've written — answers what it can, sends the rest to your inbox. Quietly nudges the project forward on its own.
Ships React, Node, Postgres. Runs tests before calling done.
Tokens, components, motion. Loads your design system.
Playwright, axe, smoke tests. Catches it before users do.
READMEs, docs, microcopy. Tone-matched to your brand.
Agents, tasks, projects, rules, memory — every piece of your setup is a simple text file in a folder you pick. Open it in any editor. Back it up. Search it. Share it.
# Fullstack Engineer ## Model: claude-opus-5 ## Description: Ships features end-to-end ## System Prompt: You are a senior engineer. Write minimal, correct code. Run tests before calling tasks done. ## Personality: Direct, pragmatic, allergic to over-engineering. ## Context: - design-system - engineering-conventions ## Allowed Tools: - Read - Edit - Write - Bash ## Connections: - github
A coordination layer that gives agents the same access you have, with the same guardrails you'd give a team.
Write your standing decisions in a plain rules file. The Chief of Staff applies them every time an agent has a question — answers what's covered, asks you for the rest. Payments and anything irreversible always come back to you.
Every task runs in its own sandbox with only the tools handed to it. Each agent's access to tools and apps is configured individually.
See each agent think and act in real time. You always know exactly what your team is doing and why.
The moment a task wraps, the Chief of Staff takes stock and lines up the next steps toward your goal — but only when there's real work to do. Momentum without the micromanaging.
Approvals, payments, new role proposals — all collected in one place. Agents wait patiently until you decide. You can also have the Chief of Staff post a daily or weekly digest to your inbox.
When agents act through your connected apps, each call is logged — who did what and whether it worked.
Run a command whenever a task moves columns. Like running tests on review or deploys on done.
For work you do more than once, save a multi-step pipeline and run it the same way every time — by hand, on a schedule, or when something happens. Approvals and stuck-agent questions land in your inbox; the rest runs itself.
BrainMocker runs entirely on your computer. It uses the Claude account you already log into — no separate API key, no extra bill, no data leaving your machine.
A native Mac app. Close the window and every background process shuts down cleanly.
Each task runs in its own sandboxed Claude Code session with only the tools BrainMocker hands it.
Your workspace is just a folder. Open it in any editor, back it up like any other folder.
Uses the Claude account you're already logged into. No new account, no extra password.
