Prefer English?
Prefer English? Switch the website language for a better experience.
On September 2, 2026, Webflow held its annual keynote in front of a room that was really waiting for one thing: to find out how far the promise of "AI that actually builds websites" would go. The answer came in four major announcements, all tied together by the same thread: turning agents into genuine collaborators within the Webflow workflow, not just writing assistants.
At Mazette.co, we have followed every Webflow Conf for years with the same lens: does this announcement actually change how we deliver client projects, or is it just product polish. This 2026 edition clearly falls into the first category. Here, straight to the point, is what was presented on stage.
Source by Webflow was the headline of the keynote. It is a new platform, separate from the historic builder, designed so marketing teams can run agents capable of creating, editing, and publishing content directly within the Webflow ecosystem, without systematically relying on a designer or developer for every micro-adjustment.
Webflow was explicit about the positioning: Source is not a replacement for the Designer, it is a complementary agentic layer, built for marketers who need speed on repetitive tasks (campaign landing pages, A/B variations, multilingual versions) without opening a ticket in the product backlog.
An important point to set expectations correctly: Source launches in a limited research preview. In other words, restricted access, reserved for a controlled number of pilot teams, with the stated goal of gathering feedback before a broader rollout. Webflow was cautious about the general availability timeline and gave no firm date.
We are dedicating a full article to Source by Webflow, its concrete use cases, and its current limitations. This deliberately concise section gives you the general framework; the details deserve their own read.
Webflow's Model Context Protocol jumped two versions at once during this conference: 2.0, presented as a structural overhaul, followed by 2.1, announced as a refinement patch arriving a few weeks later to stabilize behaviors observed in production.
In practical terms, MCP is the layer that lets external AI agents (Claude, ChatGPT, or internal agents built by agencies like Mazette.co) communicate with a Webflow project in a structured way: reading the CMS tree, understanding components, proposing changes without breaking the logic of the existing design system.
Version 2.0 introduces finer-grained permissions (an agent can be allowed to edit CMS content without touching the design), along with an enriched context system that reduces the interpretation errors seen in the previous version. 2.1 mainly fixes latency issues on projects with a high volume of components.
For an agency that builds custom automations (via Make or Zapier in particular), this maturing of MCP is bigger news than it might appear: it is the technical foundation on which any reliable agentic integration on the client side is built.
Another strong announcement: the arrival of native AEO agents, integrated into the product rather than offered as a third-party add-on. These agents are designed to continuously audit a site's structure (semantic markup, structured schemas, clarity of direct answers at the top of sections) from the standpoint of readability by AI answer engines: ChatGPT, Perplexity, Google AI Overviews.
Webflow had already laid the groundwork by opening its analytics to Claude and Gemini (we covered this in our article on Webflow AEO Analytics opening up to Claude and Gemini). The native AEO agents extend that logic: instead of simply measuring a brand's AI visibility after the fact, the agent now offers actionable fixes directly within the editor.
This is a natural evolution of what we have been documenting for several months on our own projects: classic SEO and AEO are no longer two separate disciplines, they share the same technical foundation (clean HTML structure, heading hierarchy, direct answers at the start of a section), but AEO adds a requirement for semantic clarity that these agents now partly automate.
The last major announcement of the keynote: AI Code Components. The idea is to let a designer or developer describe a custom component in natural language (a specific slider, a non-standard interaction logic) and get a generated code component that can be dropped directly into the Webflow canvas, and then edited afterward just like any native element.
This is not a simple external snippet generator meant to be copied and pasted. The difference lies in the integration: the generated component lives within the project tree, inherits the design system, and remains visually editable. For agencies that have been pushing Webflow beyond its native interactions for years with libraries like Finsweet or Finsweet Attributes, this is a clear signal: Webflow is gradually absorbing into its core product what used to belong to the third-party ecosystem.
| Announcement | Who it is for | Status as of September 2, 2026 |
|---|---|---|
| Source by Webflow | Marketing teams | Limited research preview |
| MCP 2.0 | Developers, agencies, AI integrators | Available |
| MCP 2.1 | Developers, agencies, AI integrators | Stabilization patch |
| Native AEO agents | SEO, growth, marketing | Rolling out gradually |
| AI Code Components | Designers, developers | Integrated into the Designer |
This is not a random product calendar coincidence. Webflow is responding to very real competitive pressure: generic AI tools (Cursor, Lovable, v0) are eating into the "site built by the machine" market, while legacy CMS platforms like WordPress are also accelerating their own AI layers. Simply remaining an excellent visual no-code tool is no longer enough to defend Webflow's position in the premium segment.
The strategy on display this year is therefore twofold: on one hand, open Webflow up to external agents through a robust protocol (MCP) to remain the destination platform even when creation happens through a third-party agent. On the other, build native agentic capabilities in-house (Source, AEO, AI Code Components) so as not to leave that ground entirely to outsiders. This echoes what we already observed in our analysis of the agent-focused Figma ecosystem (Figma, Motion, Shaders, Agent stuck in permanent beta): 2026 is the year design and no-code tools stop treating AI as a side feature and start weaving it into the very architecture of the product.
{{webflow}}
For an agency like Mazette.co, these announcements are not distant product theory. On a recent project for a B2B e-commerce client, we already tested a beta version of MCP to automate syncing between an external product catalog and the Webflow CMS: the time saved on product page maintenance was significant, roughly several hours a week no longer spent on manual re-entry. The 2.0 version presented at the conference, with its finer permissions, makes this kind of automation safer to deploy for a client without risking unwanted design changes.
On the AEO side, the native agents announced reinforce a practice we already apply manually on every project: structuring content to answer a question directly before expanding on it, cleanly tagging semantic hierarchy, and checking readability by AI answer engines alongside classic SEO. We break down this method in our article on SEO best practices on Webflow. Partially automating this audit does not replace editorial work, but it speeds up the diagnosis.
As for Source, its limited research preview access means widespread client adoption will take time. We recommend following its evolution closely rather than building a content strategy that depends on it starting today.
Webflow made a bold choice by betting on four structural announcements rather than a long list of cosmetic features. This is consistent with the platform's trajectory over the past two years, but it also calls for some caution: a limited research preview is still a preview, not a finished product, and MCP 2.1 being released as a quick patch to 2.0 shows that technical maturity still requires further iteration.
Our take from the field: native AEO agents and MCP 2.0 are the announcements most immediately usable on client projects today. Source and AI Code Components are promising but still need an observation period before being folded into a standardized production method.
{{seo-aeo}}
This keynote confirms a deeper trend we are watching closely: the Webflow ecosystem is no longer just a visual design tool, it is becoming an orchestration platform between humans and AI agents. For teams who want to make the most of these developments without flying blind, our method already includes continuous monitoring of these announcements, translated into concrete recommendations for each client.
If you are considering a redesign or a migration to Webflow to take advantage of these new agentic capabilities, our team supports this kind of transition every day, regardless of your starting CMS.
Code
Over 120 companies have already stepped out of the shadows with us.
Source by Webflow is a new agentic platform built for marketing teams, letting them create and edit content across the Webflow ecosystem through AI agents. It launched in a limited research preview during the September 2, 2026 keynote, available only to a small number of pilot teams.
MCP 2.0 is a structural overhaul of Webflow's AI interoperability protocol, bringing more granular permissions and richer context for agents. MCP 2.1, released a few weeks later, mainly fixes latency issues on large-scale projects rather than adding new features.
The native AEO agents announced at Webflow Conf 2026 continuously audit a site's semantic structure for readability by AI answer engines, and suggest fixes directly within the editor. They build on Webflow's earlier move to open up Webflow Analytics to Claude and Gemini.
Impactful titles, polished meta descriptions, sitemaps, Schema.org, Hn structure, and audits: these are the SEO foundations that make your content readable by both Google AND generative AI. We cover both simultaneously.
With over 180 clients, our agency is one of the few in France to hold the Webflow Premium Partner label, supporting you in every area, including custom web design, sophisticated automation, conversion optimization, and SEO and GEO visibility.
We create bespoke web designs focused on performance, user journeys, and conversion. Join over 180 clients who have trusted us to elevate their brand.
