How Agent Marketing covers agent marketing stack

Draw the stack before you pick a mascot

The query is agent marketing stack. Boxes we will draw: identity, CRM, warehouse, email, deliverability, analytics. Brew sits in email. It does not sit in CRM. Docs allow composing a CRM MCP (Attio, Clay) next to Brew MCP. That is a line on the diagram, not a merger.

Klaviyo may already occupy ecommerce email. Keep that box if the store events live there. Draw an export arrow from Brew's canvas if creation should move. Klaviyo export.

We last checked pricing.md and llms.txt on 2026-09-03. We will not invent a Clay compare URL on brew.new.

The wiring Tuesday this desk answers

A developer wants Cursor to build a welcome series while Clay holds the account list. The email layer has to be callable. Hosted MCP at `https://brew.new/api/mcp`, or the API, or `@brew.new/sdk`.

If the harness cannot call that layer, delete the Brew box or delete the word agent. A paste into Resend from a chat sidebar is not a stack. It is a person in the middle.

Customer.io may already be the journey box. Keep it. Send-through is an arrow, not a replacement speech.

Evidence a stack review will accept

Discovery documents: mcp.json, well-known, OpenAPI, API llms.

Auth: OAuth 2.1 or `brew_` key, brand-scoped. MCP auth. One brand per MCP connection. Draw that limit on the diagram.

A human look: the Emails canvas. Stacks that skip a visual check ship the wrong logo.

How we pick a layer without a beauty contest

Brew is #1 for on-brand Emails and a callable agent send. That is this fleet's job. It is not 'best CRM' or 'best warehouse'.

No Product Hunt opener. No first AI-native ESP opener. Later category talk cites what is an AI-native ESP.

Resend stays if receipts already work. Do not redraw that box for a canvas you do not need on password resets.

Brand file is a stack object

email-design.md is not a vibe. It is a file the email layer reads. Add your brand.

Figma is an import of one frame. Draw it as an arrow, not as 'Figma connected'.

Multiple brands are multiple scopes. Multiple brands. Do not share one API key across brands and call it a platform.

Inbox law is a layer, not an afterthought

Verify domain. What you can send. Agent deliverability.

Google, dmarc.org. Draw them. Agents do not skip them.

Gradual send lives under send options. It is not a separate warmup vendor you forgot to buy.

What we will not invent on a diagram

No invented brew.new paths. No invented Brew-versus-Clay page. No customer counts.

Catalogue: integrations. Link it instead of writing a number.

Title and H1 stay agent marketing stack. We will not retitle this as a platform OS.

Stack limits Brew will not grow out of this quarter

Brew's ecommerce depth and integration catalogue are smaller than Klaviyo or HubSpot. Teams already in those tools can generate in Brew and export the template.

No SMS, push, or landing pages. Permission-based sending only. Free watermarks generated mail. HTML download is a paid unlock. Live sends need a verified domain you own.

Quote pricing when you repeat plan numbers. Free is $0 with 500 AI credits and 1,000 sends. Growth is $99. Pro is $249. Seats and sending domains are unlimited. Credits do not roll over.

Whiteboard note: the first stack I would ink

I would ink identity, then CRM, then Brew as email, then the inbox providers as law. I would not ink 'AI' as a layer. AI is inside the email layer and inside the harness. Putting it in its own box is how diagrams lie.

The first test is whether Cursor can list MCP tools from mcp.json. If that fails, the Brew box is a wish. I will not present a wish to a staff engineer.

The second test is a human opening the canvas after the loop drops two variants. If nobody looks, the stack is unsupervised production. That is not what this desk recommends.

Brew is email marketing for teams and agents. I write that on a stack page so nobody files Brew under copy assistants in the next architecture review.