The In-Editor AI Assistant: Build Emails by Chatting
Build the email without leaving it
You can describe an email in plain language and watch it take shape on the canvas — without ever switching tabs, opening a separate chat window, or pasting a prompt somewhere else. Temway’s in-editor AI assistant is a chat drawer docked right inside the email and layout editors. You tell it what you want; it adds blocks, rewrites copy, restyles elements, and applies your branding, while you stay on the page and watch each change land.
It works the moment you open it — every workspace gets a real assistant on Temway’s shared AI model, on every plan, with no key to set up first. Want faster responses, no shared rate limit, or a specific provider and model? Connect your own Anthropic, OpenAI, or OpenAI-compatible key, and the assistant switches to it automatically.
For the step-by-step reference, see AI Assistant.
What the in-editor assistant actually is
The assistant is a small chat panel that opens beside the email (or layout) you’re editing. Behind the conversational front, it can do everything the visual editor can: add and reorder blocks, change their content and styling, apply your branding, and render a preview. Each action shows up in the chat as a step it took, and on the canvas as a real change to your email.
Three things define how it works:
- It works out of the box. Every workspace can chat with the assistant from the moment it’s created, on every plan — nothing to connect, no setup beyond your Temway account.
- Your own key is an upgrade, not a requirement. Bring an Anthropic, OpenAI, or OpenAI-compatible key (a local Ollama install counts) and the assistant switches to it automatically — faster responses, no shared rate limit, and the model is yours to pick. It never reaches the browser in plain text: once entered, a key is encrypted at rest and only decrypted on Temway’s server for the single moment a turn needs it. Your browser only ever holds a short, masked preview — never the key itself.
- It shares the canvas safely. The assistant takes the edit lock on the email it’s working on, so its changes and yours never collide. You can stop a response mid-stream, take the lock back, and finish a tweak by hand whenever you want.
Keys live at the workspace level by default, so one editor can add a key and every teammate’s assistant uses it. An owner or admin can instead scope a key to every workspace on the team — one Anthropic or OpenAI key covering every client an agency manages, instead of re-adding it per workspace. You can keep several keys on hand and switch between them — or between models — from the drawer.
Why it matters
- No context-switching. The fastest AI workflow is the one that doesn’t make you leave your work. With the assistant inside the editor, your email stays visible the whole time — you watch each change land, and you correct course immediately instead of describing a problem from memory in a separate window.
- Nobody hits a setup wall. A brand-new workspace can chat with the assistant immediately — there’s no “add a key before you can try this” step standing between signup and your first AI-built email.
- You can still pick your own cost and model. Connect an Anthropic, OpenAI, or OpenAI-compatible key when you want to choose the exact provider and model, skip the shared rate limit, or keep spend on your own account instead of Temway’s.
- The whole team gets it, either way. The shared model is available to everyone in the workspace from day one. If you add a key, it’s shared too — one editor connects it, every member builds with it, with no per-seat AI billing or key-passing.
- A connected key stays protected. Bring-your-own-key is only safe if the key is handled carefully. Temway never exposes the plain text to the browser after entry — it’s encrypted at rest and touched only server-side, only when needed.
How to use it
- Open the assistant. In the email or layout editor, open the assistant drawer. It’s ready to chat immediately, on Temway’s shared model — no setup needed.
- Describe what you want. “Create a welcome email from my branding — warm tone, a hero headline, a short intro, one button linking to the signup, and our social block at the bottom.” The assistant builds it on the canvas, one step at a time, and you watch each block land.
- Iterate by message. Refine in small steps — “shorten the intro to two sentences”, “give the button a larger shadow”, “add a divider above the footer”. One change per message keeps you in control and makes anything easy to undo.
- Preview, then finish. Ask for a web preview to check the result. For anything pixel-specific — a nudge to an image, an exact color — drop the drawer, take over the canvas by hand, and polish.
- Ship. When it’s ready, export the HTML or push it to your ESP for the real send.
- Optional: connect your own key. Open Settings → AI Assistant and add an Anthropic, OpenAI, or OpenAI-compatible key when you want faster responses, no shared rate limit, or a specific model. The key is validated before it’s saved, and the assistant switches to it automatically for the whole workspace.
How it differs from connecting Claude or ChatGPT
Temway supports two distinct AI surfaces, and they’re built for different moments. It’s worth knowing which is which.
Connecting Claude Desktop, claude.ai, or ChatGPT through the Model Context Protocol lets an external assistant drive Temway from outside the app — you chat in the assistant you already use, and it builds emails and layouts here through the protocol. That’s ideal when your workflow lives in that assistant, or you want to build emails as part of a larger task it’s helping with. See why MCP matters for email.
The in-editor assistant is the opposite direction: Temway hosts the chat, and it works immediately on every plan — connect your own key later if you want to. It’s ideal when you’re already in the editor and want to build or refine the email in front of you without breaking your flow. The authoring abilities are the same; the home base is different.
Many teams use both — the in-editor assistant for everyday edits and iteration, and an MCP-connected assistant for batch work or for when they’re already working inside Claude or ChatGPT. The AI vs. manual drag-and-drop article breaks down when AI authoring of either kind beats clicking.
Common mistakes
- Skipping your branding setup. “Use my branding” only works if your branding is filled in. Set your colors, fonts, and logo first and every prompt gets easier.
- Stacking five edits in one message. It works, but a single change per message is easier to review and undo. Small steps keep you in control.
- Forgetting you can stop a response. If the assistant heads the wrong way, hit stop, take the edit lock back, and redirect — you don’t have to wait for it to finish.
- Asking for content with placeholders. “Add a button” gives you a placeholder; “add a button labeled ‘Start free’ linking to temway.com/signup” gives you your button. Real copy and real links always beat “put something here.”
- Leaving an invalid key in place. If a provider rejects a key, update it in Settings — the assistant surfaces the rejection clearly, but it won’t fix the key for you.
Where to go next
- For the quick reference, see AI Assistant.
- Get copy-paste prompting recipes in Create Stunning Emails with AI.
- Understand the other AI surface in What Is the Model Context Protocol?.
- See when to lean on AI vs. the visual editor in AI Email Builder vs. Manual Drag-and-Drop.
- Avoid the common pitfalls in AI Email Writing Best Practices.
Explore: AI & MCP for Email
Building emails with AI assistants and the Model Context Protocol — from writing prompts to visual design to security best practices.