An Assistant That Draws, Not Just Writes: Visual Email Design with MCP
The ceiling of “write me an email”
Most AI email tools, if you’re generous, do one thing well: they write copy. You ask for a newsletter and you get back headlines, body text, a subject line, maybe a call to action. Useful — but when you paste it into your email builder, you’re right back where you started: staring at a blank canvas, assembling blocks by hand, picking colors, nudging spacing, trying to make it look like something a designer touched.
That’s the unspoken ceiling. The assistant handled the words; the design is still entirely on you. The result is an email that reads fine and looks like a text dump — a stack of sentences in the default font, maybe a button, no visual hierarchy, no hero, no sense that anyone composed it.
Words are the easy half. The hard half — the part that makes an email look designed — is visual. And that’s the part almost every AI tool leaves to you.
What “drawing” means here
Temway’s MCP goes past copywriting. When you describe an email to a connected assistant — Claude, ChatGPT, or any MCP-compatible client — it doesn’t just fill in text. It composes the visual layer: it places shapes, builds gradients, layers artwork front-to-back, and assembles hero sections that look designed the moment they render.
This works because the assistant isn’t emitting raw HTML or handing you a flat image. It’s calling structured visual primitives — the same ones the layout editor uses — and composing them on a free-form canvas. Absolutely-positioned text, images, buttons, shapes, and dividers, layered in the order you describe them. Gradients, shadows, corner radii, button shapes — all described by the effect you want and mapped to real settings.
The shift is real: instead of describing the email to yourself and then building it click by click, you describe the look and the assistant draws it.
The visual primitives an assistant can place
Three families of artwork are at the assistant’s fingertips:
- Shape presets. A curated library of decorative SVG shapes —
blob,burst,ring,sparkle,halo,wave-top,wave-bottom,mountains,swoop,quote, and gradient variants — that drop in without you writing a line of path data. “Scatter a few sparkles in the corners, low opacity, behind the text” is a complete instruction. - Custom vector artwork. Bespoke SVG for anything a preset can’t express — a logo mark, an icon, a custom badge. You describe or supply the markup; the assistant places it, and it’s sanitized and recolored to match your brand.
- Starter compositions. Hand-designed layout templates the assistant can apply and then refine:
gradient-hero,badge-hero,blob-hero,waves-hero,coupon-ticket,pricing-card,event-invite,stats-band,testimonial-card. Each is a full composition — background, typography, CTA, decoration — that becomes yours in one line.
The point isn’t the list. It’s that these are editable, layered vector elements — not pixels baked into a JPEG. Change the headline, swap the brand color, move the button: every piece stays live.
Content you write vs. artwork you draw
This is the mental model that makes the visual capability click.
An email is a flat, top-to-bottom stack of blocks — text, buttons, images, dividers. Great for messages, but structurally a single column. There’s no “image beside text” block; for side-by-side composition you either stack vertically or compose that piece elsewhere.
A layout is where the visual artwork lives. It’s a free-form canvas where elements sit at exact x/y positions, layered front-to-back — exactly the model you need for a hero banner, a feature section, a promo card. You design it once, publish it, then drop it into any email. When it lands in the email, it renders as a crisp image — but the source stays fully editable vector artwork the assistant can revise at any time.
So the workflow splits naturally: the assistant writes your message content (the email blocks), and it draws your hero artwork (the layout). Two surfaces, one conversation. The AI prompting guide walks through both with copy-paste recipes.
Why structured drawing beats “generate me a hero image”
It’s tempting to compare this to image-generation AI — “just have the model draw my hero banner.” But the two produce fundamentally different things, and the difference shows up the moment you need to change something.
| A generated image | An assistant-composed layout | |
|---|---|---|
| Editable afterward | No — it’s baked pixels | Yes — every element stays live |
| On-brand | Hit or miss; colors drift | Uses your branding tokens directly |
| Swap one piece | Re-generate the whole thing | Change the one element, keep the rest |
| Copy is real text | No — it’s rasterized, often garbled | Yes — selectable, crisp, spell-checkable |
| Sharpness at scale | Resizes, but stays flat | Vector — crisp at every resolution |
A generated hero image is a photograph of a design. A composed layout is the design — and you can keep editing it forever. When your brand color changes, or the CTA copy needs a tweak, or you want the same hero in three colorways, the structured artwork wins every time.
That’s the real advantage. Not “AI made a picture.” AI made a design — one that stays yours to revise.
A taste of the workflow
The full prompting playbook lives in Create Stunning Emails with AI. But the shape of a visual-drawing prompt looks like this:
Create a new layout called "Launch Hero". Apply the gradient-hero template,
then make it ours:
- Headline "Build emails at the speed of thought", white, large.
- Subheadline "Design, preview, and ship in minutes."
- Button "Start free" linking to temway.com/signup.
- Recolor the gradient from our brand color to a complementary purple.
- Scatter two sparkle presets in the corners, low opacity, behind the text.
Publish it, then start a new email from this layout with a short intro
paragraph and our footer below. Show me a preview.One prompt produced a designed hero — gradient, typography, CTA, decoration — and an email built around it. No properties panel, no pixel-nudging, no starting from blank. Describe the look; the assistant draws it.
Where to go next
- Understand why MCP matters for email more broadly.
- Get the prompting recipes — including the full layout-drawing walkthrough — in Create Stunning Emails with AI.
- Learn the layout editor and its shape, text, and image elements.
- Weigh the approaches in AI Email Builder vs. Manual Drag-and-Drop.
- Connect your assistant using the MCP documentation.