Figma is rolling out updates to its text review API, a feature of the Plugin API that gives developers finer control over text suggestions, alongside the launch of spell check in Figma and FigJam. The API supports a default text review plugin that runs automatically whenever a user types on the canvas, highlighting ranges of text and presenting a list of suggestions. Plugins built on it read that text and recommend a replacement—for correct grammar, improved copy, or alignment with a company's style guide.
Magician, a design tool for Figma built by Diagram, is one such plugin. Jordan Singer, the company's founder and CEO, participated in the API beta and described how the tool applies it. Figma Ventures, Figma's dedicated investment fund, is an investor in Diagram.
Magic spells as a product strategy
According to Singer, Magician is intended to bring AI into designer workflows and expand creativity during design—generating icons or imagery, or helping with writer's block. Each feature is called a "magic spell": Magic Icon, Magic Image, and Magic Copy. Those three are the starting set, with more planned.
The consolidated plugin was not the original plan. Magic Copy was going to ship as a standalone plugin, installable separately from Magic Icon. The team decided a single plugin holding all spells made more sense, giving it a foundation and platform extensible enough to add new spells easily and iterate on new ideas consistently.
How Magic Copy hooks into the editor
Magician uses the text review API for its Magic Copy feature, which suggests alternatives while a designer edits text layers—options for headlines, body copy, calls-to-action, and more, surfaced as the user types.
Singer characterized the API as a new paradigm for plugin interaction in Figma: rather than users working through a plugin window, the plugin runs in the background and integrates directly with the editor UI. The primary use case is presumably spell checking, but Diagram applied it to AI-powered copywriting instead.
In practice, Magic Copy targets the moment when a user has writer's block: rather than staring at an empty headline, body, or call-to-action, the designer can let Magic propose alternatives to the existing text. Singer's framing is that this speeds up design work and extends creative range.
Iteration, model choices, and guardrails
Diagram's approach centers on iteration and experimentation, including heavy fine-tuning of output for each spell so results stay consistent and useful regardless of context. The team draws on recent generative deep learning models—Singer names Stable Diffusion and OpenAI as inspirations—which he says make it faster and easier than ever to start small and experiment with AI.
The main obstacles concerned predictability and safety. Because AI output can be unpredictable, the team focused on meeting a quality bar, adding content filtering for explicit results, and testing across a wide variety of use cases.
Where Diagram sees AI in design
Diagram describes its work as exploring the untapped intersection of product design and AI; Magician is its first AI design tool and a bundle of its AI experiments. Singer says the company expects AI to have a deeper impact on product design over time, and that its aim is to build tools that make people more creative rather than replace them, applying AI to speed up designer workflows from process to design systems.
Figma credited the developer community that took part in the beta with helping thousands of users stay typo-free through their spell check plugins: @dremchee, @figm8, @hypermatic, @nowork, @sonny, @stas_sliday, @strings, and @tkmadeit.
Documentation for the text review API is available in Figma's plugin docs.



