New Webflow Designer features: the 7 announcements from Conf 2026

The seven announcements at a glance

Webflow has delivered 151 improvements since Conf 2025, almost all based on user feedback. The Builder Keynote, a new format this year, announced seven more.

‍

‍

Key takeaways: Three features are available or rolling out as of September 2, with four more arriving by the end of the year.

‍

Available since September 2

‍

Visual editing for AI code components

AI code components are used to add custom functionality without needing a dev ticket. Webflow’s example is telling: an interactive ROI calculator on a product comparison page. You describe it to the AI Assistant, and it builds the component directly on the canvas.

Until now, the only way to modify it was to re-prompt. That wasn't ideal: you’d ask for a correction, and the agent would regenerate the parts that were already working fine, too.

Now, you can click into the component, select an element, and adjust it in the Style panel just like any other Webflow element. You can also edit text directly on the canvas.

There is a third way to do it, which Webflow mentioned in passing and which is the most useful: you can drop the element you want to change into the Assistant as context, so its next pass is limited to that specific item. You no longer re-prompt the entire component; you scope the modification.

The real win isn't the Style panel. It's being able to choose between three levels of intervention depending on the scale of the change: the mouse for spacing, the canvas for text, and a scoped prompt for logic.

Available to all customers since September 2, 2026.

‍

Key takeaways: You can finally manually adjust an AI-generated component, and most importantly, refine a prompt for a single element instead of regenerating everything.

‍

Rolling out now, general availability next week

‍

Components in the CMS

Previously, inserting a custom element into a CMS item required custom code or a third-party workaround.

You can now drop a component into rich text, on the canvas, or within a CMS item, and configure it for that specific content via its props. In practice, within a rich text field, the "/" key or the "+" button opens the insertion menu; you select Component, and it is placed at the cursor.

The example Webflow provides is the perfect agency use case: you build a client testimonial component once, give it to the client's marketing team, and they add it to their new articles by updating the quote, photo, and name via the props.

But the most important point is elsewhere, and it is easy to miss. This is not limited to simple components. The ROI calculator generated as an AI code component works in the CMS just like any other component. Each instance belongs to its own CMS item, so the same calculator runs with different numbers on each comparison article, and a design change applies to all of them at once.

In other words, you can now build a feature once and deploy it across CMS items.

‍

Three limitations, as specified in the release notes:

• A component containing a Collection List cannot be placed within rich text
• A component placed within rich text cannot be bound to the CMS
• And its slots cannot be exposed

Therefore, a component intended for rich text should be designed as a closed, standalone block, configured solely by its props. You need to know this before building your library, not after.

Rolling out since September 2nd, available to everyone the following week.

‍

Key takeaway: A component, including an AI code component, can be inserted into rich text with one instance per CMS item, but it must be designed as a self-contained block configured by its props.

‍

The breakpoint canvas

Webflow has identified the core issue. A builder's day begins and ends on the canvas: managing components, refining layouts and spacing, and verifying how changes propagate across variants, breakpoints, and locales. All of this requires keeping a lot of context in mind.

There is a clear progression behind this announcement. First the dedicated component canvas, then spacing visualizations on the canvas, and finally pan and zoom. Many asked why you would need pan and zoom for a single frame. Webflow’s answer is here: because there will soon be many.

The breakpoint canvas displays all of a page's breakpoints in a single view—desktop, tablet, and various mobile formats—side by side. Make a change on desktop, and you immediately see how it propagates everywhere else, without having to click through each breakpoint individually.

The extended canvas displays desktop, tablet, mobile, and large formats side by side. The focused canvas brings you back to a single breakpoint, allowing you to work on scroll interactions or replicate a more browser-like experience.

One limitation to keep in mind: the breakpoint canvas works in Design, Build, and Edit modes, but not in Preview or Comment modes.

The benefit Webflow highlights is worth quoting directly: less guesswork, fewer forgotten overrides, and a much simpler review process, whether the work is yours, a colleague's, or something an AI agent has just changed.

Rolling out since September 2nd, available to everyone the following week.

‍

Key takeaway: All breakpoints in one view, with a focused mode for scrolling, and a review surface that works just as well for human work as it does for an agent's.

‍

Announced for the coming weeks and months

‍

Native popovers

The starting premise is spot on: every site relies on interface patterns that we rebuild over and over again. Visitors expect them, they make content readable, and they are difficult to get right, especially regarding accessibility and layout.

Webflow is set to deliver building blocks for these patterns, starting with popovers. Modals, tooltips, banners, and menus—no custom code or workarounds required.

How it's built matters. Popovers rely on modern web features, anchor positioning, and top-layer behavior instead of JavaScript. Pages remain lighter and more accessible.

And Webflow explains why: relying on native primitives allows for building properly encapsulated components. This makes popovers easier for marketers to use and makes AI agents faster to build in Webflow.

The next phase is announced for later this year: native support for all repeating elements, such as sliders, tabs, and accordions.

Rolling out in the coming weeks.

‍

Key takeaways: Native modals, tooltips, and menus using anchor positioning and top-layer instead of JavaScript, with sliders, tabs, and accordions announced for later this year.

‍

Default components for slots

This solves a problem every agency faces. A component built for one layout will eventually be dropped into a layout it was never designed for. It’s a classic side effect as a library grows and the person browsing it doesn't have the guidelines in front of them. This is even more true when an agency delivers a site to a client who doesn't know how each component was intended to be used.

Two guardrails have already been released. Visibility controls determine which components a colleague with the Marketer role can see and use in build mode, so the marketer receives a curated selection instead of the entire catalog. Slot restrictions specify which components are allowed in a given slot, ensuring a component that only makes sense in one layout stays there.

Default components add the next layer: deciding what a slot starts with. Currently, a reserved slot is empty on every new item, with no indication of what goes inside, so the same setup is repeated from scratch every time. With a default in place, the right component is already there, ready to be filled.

And here is the sentence that makes this the most underrated feature of the bunch: these guardrails apply whether the team is working directly in Webflow or agentically via the MCP. A page template therefore holds up regardless of the person, or agent, making the change.

Webflow announces availability "later this month," so sometime in September.

‍

Key takeaways: A slot starts with the right component, and visibility, restriction, and default guardrails also apply to agents using the MCP.

‍

Agentic translation

For an organization addressing multiple markets, localizing design and content is what provides a first-class experience everywhere. The real work lies in keeping the site structure aligned across locales, making every detail translatable, and adapting to the tone and cultural nuances of each market.

There is a foundation behind this announcement. Localize moved this work out of JSON files filled with strings. Then, over the past few months, Webflow synchronized the structure between primary and secondary locales, provided more control over translating component properties and custom code blocks, and made it possible to adjust tone and formality levels for each target language.

What remained manual was the translation of the content itself, even with on-demand AI. Webflow says they heard that setting up a new locale can take months.

Agentic translation changes that. Once enabled, copy updates are automatically reflected in the correct language across all locales, in real time. Thousands of clicks become a one-time setup. And teams that want to stay in control can use the translations as a first pass and refine them afterward.

We manage a bilingual FR and NL site, so we know what synchronization between two locales costs and where it goes wrong: it’s almost never the literal translation, it’s the tone that slips from one language to the other. The first-pass-then-review model is the right one. For a brand that cares about its language, we wouldn't let it run on full autopilot without a review.

Tone and formality, localized custom code, and component property controls are available to everyone now. Agentic translation is coming in the next few weeks.

‍

Key takeaways: Copy updates propagate automatically to every locale, and the first-pass-then-review mode is what we recommend for demanding brands.

‍

Style props for components

Style props are coming to components. Change a color, change a size, and limit which options are available, so that a component gives the team the right amount of creative control without opening everything up.

It’s the shortest announcement from the keynote, and it’s exactly the balance we try to strike on every delivery. Too restricted, and the client comes back to us for a color change. Too open, and the brand falls apart in six months.

Announced for later this year.

‍

Key takeaways: Customizable color and size with a limited set of options, which strikes the balance between a blocked client and a brand that drifts.

‍

Performance gains

Webflow says they have invested heavily in performance this year, both in the Designer and on published sites.

In the Designer, copying large elements is up to 98% faster and pasting is 7.7 times faster. Designer loading, page switching, locale and CMS item switching, and even typing are reported to be more responsive.

On published sites, rendering is up to 43% faster and publishing is up to 80% faster.

These figures come from Webflow and we haven't measured them ourselves. However, the slowness of copying and pasting heavy components has been a long-standing complaint, so we'll take it. Webflow also notes that this is just the beginning of a longer investment roadmap.

Live for all customers.

‍

Key takeaways: Up to 98% faster copying and 7.7x faster pasting in the Designer, 43% faster rendering, and 80% faster publishing on live sites.

‍

The foundation

Three of the seven announcements aren't isolated new features; they are layers built upon work delivered in recent months. Webflow says this itself, and it matters when assessing their reliability.

Spacing displayed on the canvas instead of just the Style panel, the dedicated component canvas with side-by-side variants, and panning and zooming on the page. The breakpoint canvas is built on these three.

Component visibility controls and slot restrictions. Default components pick up exactly where those leave off.

Localize improvements: tone and formality by target language, localized custom code in the head and body, and more control over translating component properties in secondary locales. Agentic translation is built on this.

Which means something practical: these three features are the most predictable of the bunch because they extend something that is already running.

‍

Key takeaways: The breakpoint canvas, default components, and agentic translation extend previously delivered work, making them the most reliable building blocks of the lot.

‍

The thread Webflow repeats three times

These announcements are presented as builder convenience. They are something else, and Webflow says so three times without dwelling on it.

On popovers: relying on native primitives allows for properly encapsulated components, which makes AI agents faster to build in Webflow.

Regarding default components for slots: these guardrails apply whether the team is working directly in Webflow or agentically via the MCP.

Regarding the breakpoint canvas: the release note mentions making it easier to review AI-generated changes across all devices.

Side by side, these three sentences tell the same story. An encapsulated component is one that an agent builds without error. A restricted slot is a slot where an agent cannot place just anything. A canvas that shows all breakpoints is a review surface for work that you didn't do yourself.

It’s the same logic as the context layer Webflow mentions regarding Source: encoding rules in one place so that agents work within them. The Designer is becoming that place.

‍

Key takeaway: Encapsulated popovers, constrained slots, and review canvases aren't just for convenience; they are the foundation that makes agentic work controllable and reviewable.

‍

What this changes for an agency

Components in the CMS solve the most painful delivery problem. No more clients asking for help to insert a block into an article: with a component and props, they can handle it themselves.

The three guardrails combined—visibility, slot restrictions, and default components—constitute a new type of deliverable. It’s no longer just "here is your site," it’s "here is your site and here is what your team can do within it without breaking it." That can be documented and billed for.

The breakpoint canvas saves time every single day, no question about it.

And a constraint to integrate before building, not after. The limitations of components in rich text change how you design your library: no Collection Lists, no CMS bindings, no exposed slots. Components intended for rich text form a separate family, closed and configured by their props.

‍

Key takeaway: Components in the CMS eliminate back-and-forth publishing, and the three guardrails become a deliverable that can be documented and billed.

‍

What this changes for a marketing team

Three things are usable right now, or nearly so.

Visual editing of AI code components: a spacing issue can be fixed by hand, without re-prompting and without the risk of breaking what was already working.

CMS components: a rich block can be inserted into an article without needing any help.

Guardrails: what you see in the component catalog is what works where you place it. Fewer choices, fewer errors.

And agentic translation if you are multilingual, with the caveat mentioned above. For a first pass followed by a review, yes. For full automation, it depends on your standards for language quality.

‍

Key takeaways: Fix an AI component without re-prompting, insert a block into an article independently, and use a filtered component catalog that only suggests what actually works.

‍

Our take, feature by feature

CMS components: the feature that changes the most on a daily basis, and the combination of AI code components with per-item CMS instances simplifies projects that used to be complex.

Breakpoint canvas: the most immediate time-saver, and the focused scroll mode avoids the trap of a single-view perspective.

Default components for slots: the most underrated feature, because it is the only one that imposes constraints on an agent.

Visual editing for AI code components: the real gain is the prompt scoped to a specific element, not the Style panel.

Popovers: top-layer and anchor positioning are things we currently code by hand, and default accessibility is the strongest selling point.

Agentic translation: the right model is a first pass, not full automation.

Style props: concise, long-awaited, and exactly the right level of control.