An IDE for entrepreneurs to build digital products on their own, at the pace of a whole team— from the first prompt to the product, from the product to the sale, with a team of AI agents that grows with your business.

  • Runs on the Claude Code, Codex and Antigravity you already pay for
  • From US$ 10/mo · 7 days free, no card

macOS (Apple Silicon) · Windows

CODING / Workspace 1 / Grid
3 agents talkinggrid · 3 panes · maestro orchestrates~/products/brix

Connect several agents, name a leader and let the team run your project in parallel — from code to ads, carousels, content and video. A complete environment, with 13 exclusive skills, MCP and agents with persistent memory.

The problem

The problem isn't the AI.
It's the mess around it.

You already vibe code. You have Claude Code or Codex open, you know how to ask and how to follow up.

But the product doesn't end at the code. It ends at the sale. And between the two, your day turns into this:

  • One terminal per agent — and none of them knows what the other did.
  • The agent stopped to ask for permission twenty minutes ago. You were in another tab.
  • ChatGPT writing the copy from scratch, because it has no idea what you built.
  • Canva for the creative. CapCut for the video. Ads Manager to publish. A spreadsheet to measure.
  • And tomorrow, explaining the product all over again to every tool — because none of them remembers.

Eight tools. Zero shared context. You're not building a business; you're juggling.

The problem was never the intelligence of the models.
It was the environment around them.

What BRIX is

The IDE that keeps going
after the code.

Every IDE stops at the code. BRIX is the ADE — Agent Development Environment: the place where an entrepreneur — developer or not — builds the product and its marketing in one place, with a team of AI agents they conduct.

The whole cycle, in a single folder:

  1. CODINGYou build the product with a team of agents.
  2. COPYThe AI reads the product base and writes the offer.
  3. STATIC ADSThe creatives come out ready, in batches.
  4. VIDEOCuts, captions and motion in HyperFrames.
  5. META ADSPublish, measure and go back to the start.

Every studio reads the same product base in Markdown, in your folder. Switching studios changes the output format — never the context. That's why the ad knows what the sales page promised.

Code tools stop at the first box. Marketing tools don't have the first one. BRIX runs the cycle end to end.

How it works

Three mechanisms, none of them decorative.

1

Orchestration: you conduct, the agents type

One command and the AI splits into several terminals that collaborate. You connect one agent to another with an edge on the canvas — the edge is the permission: once connected, they exchange prompts, read each other, split the task and deliver together.

Runs withClaude CodeCodexAntigravityOpenCodeyour usual shellon the subscriptions and tokens you already pay for — BRIX doesn't charge for the model

Maestro and ready-made squads

Maestro: name a leader; it recruits the workers, distributes the work and follows up on delivery. Squads: teams with roles, briefing and connections already drawn — feature, cacabug, blindagem, oferta, conteudo — plus the ones you create.

Grid or canvas

In the grid, panes rearrange themselves with every new terminal. On the canvas, you position, zoom, pin notes, task lists and the images the agent published.

Scene

You watch the team at work.

Each agent becomes a pixel-art character that walks to a desk and works according to the terminal's real activity — reading, writing, using a tool or waiting for you. It isn't an animation: it's the actual terminal.

2

One product base every studio reads

The product library is local, in Markdown, Obsidian-style: product.md, audience, offer, brand, FAQ. Every studio — copy, ads, carousel, video — reads those files before producing. The copy comes out specific because the AI read your product; the creative repeats the page's promise because it drank from the same source.

3

Agents with persistent memory

A regular terminal starts empty every session. A BRIX SuperAgent is yours: it has a personality, rules, a knowledge base (PDF, DOCX, MD) and a memory that grows with what you teach it. What you corrected today, it remembers next session, next week, on the next product. Each product also builds up a living copy memory: what worked and what not to repeat.

The engine is per agent: your ChatGPT subscription or OpenRouter for any other model — DeepSeek, Kimi, GLM, Claude, Gemini.

That's why the environment is worth more every month — not because you pay more, but because it knows you better.

BRIX Squads — on every plan

Say the goal.
The team is born ready.

A squad is a team of agents that is born ready: each one with its role, its briefing, and already connected to whoever it needs to talk to. One click and the agents show up in the Workspace, pre-programmed for your goal. Five ready-made — code and marketing — or build your own, with whatever roles and connections you want.

BRIX Squads

End-to-end feature

Architect, Backend, Frontend and QA building a whole feature

4 agents · 6 connections
Who's who
  • ArchitectClaude Codedesigns the solution and coordinates
  • BackendClaude Codeserver, data and routes
  • FrontendClaude Codeinterface consuming the contract
  • QACodextests and reports to whoever broke it
Duplicate and editCreate squad (4 agents)
  • 1 clickand the whole team shows up in the Workspace, briefed and connected
  • 5 ready-madeFeature, Bug hunt, Hardening, Offer and Content
  • your ownroles, briefing and connections however you want — or by command: medmaestri squad
Maestro — on the PRO plan

Say what you want.
Maestro runs the app.

Maestro is the BRIX voice agent. You say “Hey, BRIX” — or hold ⌘⇧O — and ask: a squad, a prompt for an agent, how the other one is doing. It acts through the same tools the BRIX MCP exposes, answers with a neural voice and lets you know when someone has finished or is waiting on you. Runs on the model and voice you choose, with the day’s cost on screen.

Maestro · LiveSetup
Go ahead, I’m listeningSent automatically when you stop talking.
YouHey, BRIX — create a bug-hunt squad and put it on the checkout bug
MaestroCreated Repro, Investigator and Fixer, connected. Sent the briefing. I’ll let you know when Fixer is done.
Actionbrix_squad cacabug ✓ 3 agents
MaestroFixer is done: the bug was the coupon with a trailing space. Want me to ask for the test?
Write to Maestro…
Today US$ 0.08 · View usageDeepSeek V4.1 Flash · Ara

You talk, it acts

The same tools as the BRIX MCP: creates terminals, teams and squads, sends prompts, asks how things are going, switches mode, writes notes and tasks, connects one agent to another, opens the browser.

  • "create a bug-hunt squad"
  • "have Beto review the offer"
  • "how is Ana doing?"
  • "connect Reviewer to Backend"

It reads your project

Sees the folders of the open Workspaces and the active product: summarizes CLAUDE.md, design.md, product.md and the offer, and reads a file or lists a folder when you ask. Only inside those roots.

It lets you know

When an agent finishes or is waiting on you, it speaks — at most one notice every 20 seconds, grouped. Off by default; turn it on in Setup if you want.

Your voice and model

Runs through OpenRouter, with your key. Model and neural voice are chosen in Setup, with "Listen" on each voice; the cost of every turn is added up per day and stays on screen.

⌘⇧O tap opens the drawer, hold to talk · ⌘⇧U passive listening “Hey, BRIX”

The arsenal — PRO plan

A pack of 13 skills, an exclusive MCP,
five studios.

13 marketing skills installed straight into the agents you already use. Copywriter, designer, video editor, social media and media buyer — each one is a studio, and they all read the same files of your product.

STATIC ADS

43 templates already proven to convert: WhatsApp chat, iPhone notification, newspaper headline, PIX receipt, five-star review, Reddit thread. 45 creative archetypes to stop the scroll, 15 variation recipes in one click. The queue runs in parallel without blowing past the API limit, and every ad is born with the name your team uses.

  • 43 templates
  • 45 archetypes
  • 15 variations
  • parallel queue

COPY

Market diagnosis, sales angles in a 3D constellation, copy by goal (sales page, ads, email, VSL, headlines, bullets) and audit. With the product’s living memory.

CAROUSEL & CONTENT

12 voices and formats, from Storyteller to Contrarian, as Reels, carousel, thread, post or newsletter. HTML preview, PNG export, Obsidian-style writing with a file tree and live reader.

VIDEO

An editing suite driven by natural language on HyperFrames: cuts pauses by voice detection, transcribes and captions, does rough cuts and motion — the queue processes each video and shows the preview. Plus 4 AI video-generation models, with the cost on screen before you confirm.

META ADS

MEDCODING STUDENTS

The dashboard of your campaigns, fed by an agent connected to Facebook’s official MCP. Spend, ROAS and CPA — and you talk to it to understand and act, without leaving BRIX. Not for sale here: exclusive to MEDCODING students.

BRIX MCP

The app’s exclusive MCP: run BRIX from outside, through Claude Desktop or Codex. Create squads, send prompts, read replies, publish notes — by command, without clicking anything.

And also, on every plan

Guard, the auditor

The agent that sweeps the repository for vulnerabilities and delivers the report by severity on the canvas.

Built-in browser

Inspector, device emulation and DevTools inside the app.

Voice both ways

Dictation in the terminal and Call Mode, with recognition 100% on-device — nothing goes anywhere.

Code, .env and review

Code editor, form-style .env editor and diff review with commit and push.

SuperAgents — on every plan

SuperAgents: agents with persistent memory
that learn from you and work on your team.

A regular terminal is born empty every session. A SuperAgent is yours: it has a name, personality, rules, knowledge base (PDF, DOCX, MD), its own skills and a memory that grows with what you teach it. What you corrected today, it remembers next session, next week, on the next product — and it talks to the other agents on the canvas by name.

Lia

Program copywriter

Style
Direct
Engine
OpenRouter · DeepSeek V4.1 Flash
Skills
3
Memory
7 entries · 612/2200
Created
Sep 2

Theo

Carousel designer

Style
Didactic
Engine
ChatGPT (subscription) · GPT-5.6 Luna
Skills
1
Memory
2 entries · 156/2200
Created
Sep 11

Bia

Code reviewer

Style
Executive
Engine
OpenRouter · Kimi K2
Skills
4
Memory
11 entries · 1340/2200
Created
Aug 19

Otto

Offer strategist

Style
Executive
Engine
OpenRouter · Claude Sonnet 5
Skills
2
Memory
5 entries · 488/2200
Created
Aug 27

The engine is per agent

Runs on your ChatGPT subscription (login, no key) or through OpenRouter, which opens up any other model — especially DeepSeek, Kimi and GLM. Also takes a direct key from OpenAI, Anthropic and Groq.

Joins the grid

Opens as a pane next to the terminals, in the Grid or on the Canvas. Takes an edge, shows up in the Scene and shows when it’s working — like any other agent.

Talks by name

The other agents talk to it with ask, send and check, the way they talk to each other. Claude Code asks Lia what the product is called — and she knows.

In practice

What changes in your day.

  • How three agents connected by an edge do in an afternoon what one chat window does in a week — and why the difference isn't the model, it's the connection.
  • Why ChatGPT's copy comes out generic — and the 40-line file that fixes it for good.
  • Why a cacabug squad finds in minutes the bug you were going to hunt after lunch.
  • What happens when the ad and the sales page read the same oferta.md (spoiler: they stop contradicting each other).
  • How to fire off 6 creatives at once without blowing up the API — and get them back already named the way your team files them.
  • The question the SuperAgent will never ask you again after the second session.
  • How to know, without opening the terminal, whether the agent is working or waiting for you — by looking at a pixel-art office.
  • The .env editor that treats environment variables like a form — and puts an end to "I forgot to swap the key".
  • Why you can close BRIX, open the folder in Finder and find everything in Markdown, PNG and MP4 — versionable in git.
  • The command that creates a whole team — supervisor + workers, already briefed and connected — before you finish your coffee.
Who it's for

For those who build. For those who sell. For those who do both.

Developers who run a business

You already pay for Claude Code or Codex and feel you use a fraction of it: one window at a time, several terminals by hand, and the product's marketing gets left for later. BRIX is where the agents work together — and where "later" happens in the same place.

Solo builders who also sell

Indie hacker, solo founder, digital product creator. You need a dev, a copywriter, a designer, an editor and a media buyer, and you have no team and don't want an agency. BRIX PRO is that team, on your machine.

Marketers who want to use AI for real

No coding: you pick the product, fill in the base and the studios do the work. Copy, creatives, carousel and video with the context of YOUR product — not the chat's generic one.

Requirement: macOS with Apple Silicon (M1 onward) or Windows.

Plans

Two doors, the same engine.

The difference is how much of the arsenal you unlock. STARTER isn't a sample: it's the whole agent environment.

BRIX STARTERCODING MODEFor those who want to build with AIUS$ 10/mofirst 3 months · then US$ 14/mo

The complete ADE. STARTER isn't a sample: it's the whole agent environment.

Subscribe to STARTER
  • Orchestration terminal in a grid or free canvas
  • Maestro mode — one command, several terminals collaborating
  • Ready-made Squads with roles, briefing and connections
  • SuperAgents with persistent memory, knowledge base and skills — on your ChatGPT subscription or via OpenRouter (DeepSeek, Kimi, GLM…)
  • Scene view — the pixel-art office reading real activity
  • Built-in browser with inspector, device emulation and DevTools
  • Voice dictation and Call Mode, 100% on-device
  • Code editor, .env editor and diff review with commit and push
  • Guard — the security auditor
  • BRIX MCP — operate the app from outside
the most complete for anyone who wants to sell a product
BRIX PROCODING MODEMARKETING MODEFor those who build and sellUS$ 30/mo

Everything in STARTER, plus the marketing arsenal: copywriter, designer, editor and social media in one app.

Subscribe to PROEverything in STARTER, plus
  • COPY studio — local library with Obsidian-style RAG, angles in a 3D constellation, copy by objective, living memory per product
  • Static Ads — 43 templates, 45 archetypes, 15 variation recipes, parallel queue
  • Carousel & Content — 12 voices and formats, HTML preview, PNG export
  • Video — natural-language editing in HyperFrames (silence cuts, captions, motion) + 4 AI generation models with the cost shown before generating
  • Pack of 13 marketing skills installed on your agents
  • Maestro — the voice agent that operates the app (⌘⇧O)

Want to try it first? Download the app and create your account: 7 days of STARTER free, no card.

The math that matters

Connect the subscriptions you already have.

Each agent runs on the subscriptions and tokens you already pay for — Claude Code, Codex, Antigravity, ChatGPT or OpenRouter. They stay yours, paid directly to the provider. The BRIX subscription is for the environment — the terminals, the canvas, the maestro, the studios and the skills. Do the math the way it actually is:

  • US$ 0.33per daySTARTER: US$ 10 per month for the first 3 months, then US$ 14. Less than a coffee for the environment where the agents you already pay for work together, instead of one at a time.
  • US$ 1.00per dayPRO: US$ 30 per month. For what you pay separately today — the design app for creatives, the video editor, the copy tool, the carousel generator — or for the hourly rate of each professional who does it for you. A single freelancer briefing costs more than the whole month.
  • US$ 0markupNo double billing. BRIX doesn't charge for tokens, per agent or per creative generated. What goes up is the usage of the subscriptions you connected — the same it would with a single window.

If BRIX saves you an hour a week, it paid for the month. If you vibe code and market your own product, it saves you that per day.

META ADS (the campaign dashboard via Facebook's official MCP) isn't sold here — it's exclusive to MEDCODING students.

Proof — what you can verify before paying

We won't ask you to take our word.
Download and look.

  1. Open the Scene and watch the character stop when the terminal stops. The reading comes from the PTY, not a timer.

  2. Open the product folder in Finder: product.md, publico.md, oferta.md. That's what the AI reads. Edit it by hand if you like.

  3. Count the templates in Static Ads: 43. The archetypes: 45. The recipes: 15.

  4. Close BRIX and open what it produced without it: Markdown, PNG, MP4, JSON. Version it in git. Take it with you.

  5. Read Apple's signature on every update: signed and notarized, macOS opens it without complaint.

Guarantee

Two safety nets, zero red tape.

The risk of trying is zero. The risk of not trying is one more month paying for the best agents in the world to use one window at a time.

Before paying7 days of STARTER free, no card.

Download the app, create your account and use it.

After subscribing7-day guarantee, no red tape.

Didn't become your environment? We refund you.

What you create never leaves your machine.

The product base, the copy, the creatives, the videos and the campaign history live in files on your disk — Markdown, PNG, MP4, JSON. You can open all of it without BRIX, version it in git, take it with you. Login validates access, and that's all.

Why now

There's no countdown on this page.

What there is, is arithmetic: every month with a single window is a month paying for Claude Code or Codex for a fraction of what it would do with three connected agents, a canvas and a studio reading your product.

And every month without memory is a month explaining your business from scratch — to an AI that will forget it tomorrow.

The environment that remembers gets more valuable every week you use it. The sooner you start, the more it knows.

Download for macOS — 7 days free, no card
Before you ask

FAQ

Will I pay for the models again?

No. BRIX uses the CLIs already on your machine — Claude Code, Codex, Antigravity, OpenCode — with the login you already have. Each agent runs on the subscriptions and tokens you already pay for. The subscription is for the environment.

And the SuperAgents — which model do they run on?

Whichever you choose: your ChatGPT subscription (login, no key) or an OpenRouter key, which opens the whole catalog — DeepSeek, Kimi, GLM, Claude, GPT, Gemini. It also accepts direct keys from OpenAI, Anthropic and Groq. The model is per agent: one can be DeepSeek, another GPT.

Do I need to know how to code?

For the marketing studios, no: you pick the product, fill in the base and the AI works. CODING is a real terminal — but the agents are the ones typing the commands. You conduct.

I already use Cursor, Warp or Claude Code straight in the terminal. Why another environment?

Because they stop at the code. Cursor and Warp are great for writing and running — and BRIX runs alongside them, not instead of them: your Claude Code and your Codex are still the agents. What none of them has is the rest of the cycle: several agents connected by edges on a canvas, a maestro that distributes work, memory that persists across sessions — and the studios that take the same product and turn it into copy, ads, carousels and video. Code tools stop at the first box; marketing tools don't have the first one.

Does BRIX replace Claude Code?

No, it conducts Claude Code. It adds what's missing when you have a single window: several agents at once, connected to each other, with a canvas where they publish notes, tasks, images and reports — and the studios that turn the same product into marketing.

Does my data go to a server?

No. Everything you produce stays in files on your disk — Markdown, PNG, MP4, JSON. Login exists only to validate access.

Does it run on my computer?

macOS with Apple Silicon (M1 onward) and Windows.

What's the difference between STARTER and PRO?

Both doors lead to the same engine. STARTER (US$ 10 for the first 3 months, then US$ 14) is the whole ADE. PRO (US$ 30) unlocks the marketing arsenal — COPY, Static Ads, Carousel & Content, Video, the pack of 13 skills — and Maestro.

What about updates?

The app updates itself. Every version is signed and notarized by Apple.

Can I try it first?

Yes: 7 days of STARTER free, no card. After subscribing, a 7-day guarantee.

An army of one:
you and your agents.

Dev, copywriter, designer, editor and media buyer working in parallel, at your pace, with the context of your product. Stop juggling eight tools: connect the agents, name the leader and let the team run the project — from the first prompt to the product, from the product to the sale.

  • 20agents in one Workspace, each on its own task
  • 1person conducting — you
  • 0times explaining the product again

From US$ 10/mo · 7 days free, no card · macOS and Windows

You conduct. The agents type.