Telex is an experimental, browser-based free tool from Automattic that uses AI to generate WordPress blocks from natural language prompts.

Instead of setting up a local development environment, configuring build tools, and scaffolding a block by hand, Telex lets you describe what you want, preview the result, and download a ready-to-install plugin.

This article explores what that workflow looks like in practice and what kinds of blocks it is realistically suited for today. Using two real examples, including a more complex editor-only block, we look at how far Telex can take you and where manual intervention is still required.

Telex user interface
Telex user interface.

How Telex works in practice

Telex follows an assembly-line style workflow. You describe the block you want using natural language, then Telex maps that request onto a predefined WordPress block scaffold based on the standard create-block tooling.

How Telex works
How Telex works.

From there, AI fills in the details such as block metadata, editor UI, rendering logic, and basic styling within those constraints. The block is then built in a hosted Node.js environment, previewed in WordPress Playground, and packaged as a ZIP plugin that’s ready to upload.

Under the hood, the generated plugin follows a familiar structure, closely aligned with create-block:

block-name/
├── build/
├── src/
│   ├── edit.js
│   ├── save.js
│   ├── index.js
│   ├── style.scss
│   └── editor.scss
├── block.json
├── package.json
├── readme.md
└── .gitignore

Let’s build our first of two blocks.

Vibe coding an Embed ChatGPT block

You may recall a time when it was common to see a Google search box embedded in a web page, allowing users to search a site or the web. If you need to provide that kind of experience for your users, but this time with ChatGPT input, here’s how.

This is a great place to start, as you can focus on getting comfortable with Telex’s UI without being too concerned about prompt refinement or multiple iterations.

Start as a logged-in user with wordPress.com credentials, and you can describe your initial prompt. Here is what we used:

We would like to create a block that embeds the ChatGPT prompt to allow users to use it on the frontend.

See the UI below:

Telex interface showing an input prompt for configuring an automation.
Initial Telex input prompt.

Telex creates the ChatGPT embed block, which includes the prompt input box, message history display, and styling that matches ChatGPT’s aesthetic.

At this point, the block outputs an error message in the editor UI.

Block displaying an error message during workflow execution.
Block displaying error message.

We asked Telex to resolve the issue and were guided through generating a ChatGPT API key and adding it to the block via the Inspector sidebar. In this case, we felt that using an external API for the block did not introduce an unwarranted security risk.

Block inspector panel showing configuration options and settings.
ChatGPT API Key added to the Block Inspector.

Finally, we ran a quick test on the frontend to confirm that everything was working as expected.

Frontend displaying the initial response generated from a prompt.
Frontend displaying initial prompt and response.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.