How does OptiMonk's AI agent build and publish a popup?

Updated

OptiMonk's AI agent turns a plain-language request into a live, A/B-tested popup campaign: you describe the popup you want, the agent designs it, writes the copy, and puts it live on your store. It can also attach an image, set a coupon code, and choose when the popup appears. This is aimed at store owners who want a campaign running without briefing a designer or writing code — OptiMonk states it is free to start with no credit card required.

What you give the agent

The input is a description, not a form. On the site, the prompt is simply "Describe the popup you want," with an option to attach an image. You can also name the store and the offer, as in the example shown on the page: create a welcome popup for allbirds.com with 10% off the first order.

Common popup types the agent supports include:

  • Welcome popup
  • Exit-intent discount
  • Spin-to-win
  • Back-in-stock and cart-saver campaigns (shown as recommended opportunities)

How the agent builds it

The agent works in a conversation and reports each step back to you. In the Allbirds example on OptiMonk's page, the flow looks like this:

  1. It reads the brand. The agent says it read the Allbirds brand and drew four concepts — Natural Step, Wool Wonder, The Perfect Pair Quiz, and Eco Reveal.
  2. You pick a direction. The user replies "Go with A," and the agent proceeds with that variant.
  3. It sets the campaign settings. The user specifies "show it to first-time visitors after 5 seconds," and the agent applies that targeting and timing.
  4. It sets the coupon. The user supplies code WELCOME10, and the agent attaches it to the campaign.

Each of these appears as a tool action in the conversation (for example, select_variant, set_campaign_settings, set_coupon), so you can see what the agent changed before anything goes live.

Publishing and what happens after

Once the design and settings are confirmed, the agent puts the campaign live on your store. From there, OptiMonk says the same conversation continues to manage the campaign rather than stopping at publish:

  • It tests itself. The agent runs A/B tests and learns from real visitors. The page shows a Control at 7.6% conversion versus a winning Variant at 12.1% — a +59% uplift with a 97% chance to win — then asks whether to roll the winner out to everyone.
  • It animates. One tap turns a still photo of a finished popup into a looping video background.
  • It finds the next campaign. The agent surfaces new opportunities (for example, a cart saver for a high-bounce landing page or a back-in-stock popup for a frequently sold-out product) and keeps them ready for you to build.

Building from the AI tools you already use

OptiMonk also offers an MCP server so you can build popups from Claude or ChatGPT and run the campaign in a chat, with no code needed. The page shows a Claude conversation using OptiMonk tools to design, select a variant, set campaign settings, and set a coupon — the same steps as the in-product agent, just driven from an external assistant. OptiMonk provides a "Connect OptiMonk" option and documentation for this setup.

What to check before you start

  • Store access. The agent publishes to your store, so you need OptiMonk connected to the store you want the popup on.
  • Offer details. Have the discount or incentive ready (for example, a coupon code) so the agent can attach it in the same conversation.
  • Targeting. Decide who should see it and when — the example used first-time visitors after 5 seconds — since the agent applies these as campaign settings.
  • Pricing and plan limits. OptiMonk links to a pricing page, but the specifics of paid tiers are not detailed in the available page content, so check that page for current terms before committing to a plan.