Feature Requests

Redirects: inserting a RedirectNode fails with “Parent … cannot contain this target”. Custom code: no way to write site-wide head or body code. Localization: no locales or translations. Local files: no way to read from disk or export to it. Managed CMS collections: can’t create or sync them. Deployments: no deployment history, issues or domain selection. Code files: no version history or renaming. Docs: many api methods aren’t documented.

With Framer MCP, AI agents can build most of a site. But on real projects with code components, they hit walls that only a person in the editor can get past. Every item below hit us on a live store: Code components are only half writable Fonts: font controls can't be set, and custom fonts aren't available to plugins at all, so every agent-built text falls back to Inter. Grouped properties: Object controls (text styles, box styles, layout groups) are rejected. New properties after a library release: reported as "Unknown control" by the agent, while the editor shows them. Instance updates: library instances can only be updated by clicking Update in the editor, one by one. Slots: slot values can't be read through the plugin, aren't passed to hidden(), and inserting an instance into some components fails. Editor data: CMS link lists arrive empty on the canvas and in Preview, and components with an auto layout get no width. Branching and publishing aren't safe for agents One shared branch: all agent sessions on a project share one active branch, so two agents switch each other's work. Creating branches: branches can't be created on demand, and switching then editing can land on the wrong branch. Publish preview: on main, it publishes production. Publish diff: it lists component names, not what changed, so an agent can't judge whether a change is safe to ship. Agents can't see the result Preview: no access to the editor's Preview or its console. Library versions: no way to see which version an instance is on versus the latest. Compile errors: a code file that fails to compile gives no clear error back. Ask: what the editor can do, the Plugin API and MCP should be able to do too. That means writable fonts (custom included) and grouped controls, a programmatic instance update, branches per session, a publish preview that doesn't deploy, a readable diff, and access to Preview and its errors. Agents can already build the structure of a site. With these, they could finish and ship it without anyone fixing it by hand.

Add a Server API method that returns the pages changed in a published deployment, or between two deployments. The Staging & Versions panel already shows this per version under "Changes" (page paths, "Published by", "Published at"), so the data exists; it is just not reachable from the API. Suggested shape (any equivalent is fine): framer.getDeploymentChanges(deploymentId: string): Promise<{ added: string[]; removed: string[]; modified: string[] }> // or framer.getChangedPaths({ from: deploymentId, to: deploymentId }) Why we need it We run a security scan before promoting a staging version to the custom domain. The scan must cover every page changed since the deployment currently connected to production. Today that list has to be typed by hand from the Versions panel, which is the one error-prone manual step left in an otherwise automated flow (Slack workflow → Server API → scan → approval → deploy()). What we tried with framer-api 5.1.0 getUnpublishedPageChanges() / getChangedPaths(): only unpublished changes since the last publish. Empty right after a designer publishes to staging, so the change list of that version is lost. listDeployments() / getDeployment(id): id, deployedBy, createdAt, updatedAt, versionUrl, status only. getChangeContributors(fromVersion, toVersion): returns user ids, and the version numbers it expects are not exposed anywhere we could find. Diffing the sitemaps of two version URLs: catches added and removed pages, not edited ones. Diffing rendered HTML of two version URLs: unusable, since a Framer runtime update changes every page (884 of 2048 pages differed with no content change). Framer support confirmed on 2026-10-28 that no such API exists and that there is no public roadmap date. Context Project: 2-0-New-Web (staging enabled, custom domain opensurvey.io ) Volume: roughly 10–25 production deploys per month, often 10–20 stagins Use case: pre-deploy security scan (OWASP checks via Playwright) on changed pages only; deployedBy and createdAt from listDeployments() are already enough for the "who/when" part, so the only missing piece is "which pages". Thanks for considering it.

I've been using the Framer integration with ChatGPT to make updates to my portfolio website and have run into an issue where changes made through the integration appear correctly in the Framer editor but aren't reflected in Preview or on the published site. The changes are clearly visible in the editor, but Framer continues to show "No changes to publish." Even manually publishing doesn't push those updates live. I've tried refreshing the editor, testing in Incognito, and republishing, but the issue persists. I also tested making changes directly in the Framer editor, and those work as expected. Manual edits trigger the unpublished changes notification and appear correctly in Preview and on the published site. The issue seems isolated to changes made through the ChatGPT integration. I also checked the branch menu as suggested by Framer Support and confirmed I'm working on the main branch, with no other branches available. Expected behavior: Changes made through the ChatGPT integration should be recognized as unpublished changes and reflected in Preview and on the published site after publishing. Actual behavior: Changes appear in the editor but aren't recognized by Framer's publishing system or reflected in Preview. Project: https://framer.com/projects/YTi3FqcwYGiHJoTk3Dy1 Published site: https://annabuchy.com

Plugins that wrap a third-party runtime (a rendering SDK, a player, a shader library) must bundle that runtime and pin it to an exact version. That's a sensible rule: code on a published site shouldn't change after review. The catch is that a pinned component never changes at all. When the library vendor ships a fix, it reaches new insertions only after the plugin releases an update and it clears review. Every component already placed keeps the old version until each user finds it and re-inserts it by hand, and most never will. If the library had a security hole, it would stay live on published sites indefinitely, however quickly the vendor and the plugin developer reacted. Any of these would close the gap while keeping review in the loop: Let a plugin publish a reviewed update to a module it hosts, which placed components pick up automatically within a compatible version range. Show an update notice on placed components whose module has a newer reviewed version, with one click to update every instance in the project. For plugins that are the official integration of a vendor, allow loading the vendor's own library from the vendor's domain, pinned or verified by hash. This affects any plugin that inserts code components built on an external library, not just one plugin.

Please add a native Framer connector/plugin for normal ChatGPT conversations, not only Codex or external agents. I would like to authorize ChatGPT to access and edit a specific Framer project directly from a normal ChatGPT conversation, with permission controls for what ChatGPT can change.

We just launched a small brand on Framer using Shopify and were quite surprised to realise that we needed a third paid subscription just to properly connect the two. We’re already paying for Framer + Shopify, so having to add another monthly subscription for a third-party plugin is quite a lot for a small brand just starting out — and it wasn’t something that felt very clear before launching. It would be amazing to have a native Shopify integration directly in Framer, at least on paid plans, with product sync, add to cart, cart and checkout. We love Framer for the creative freedom it gives us, and I think this would make it a much more complete solution for small ecommerce brands.

Currently only OpenAI and Anthropic models are available. I think (though the models are lacking behind a bit) it would be good to add Mistral's models to Framer considering they are the only European AI company with their own LLM. Framer is a great app that is also European which automatically makes it more interesting for companies striving for European sovereignty. Having the option of using a European AI model would just be that much better.

We need to add a floating accessibility button to our website to ensure compliance with accessibility regulations and WCAG 2.1 Level AA standards. We are looking for a standard widget like the ones commonly seen on accessible websites that visitors can simply click to activate features such as text resizing, high contrast mode, keyboard navigation, and dyslexia-friendly fonts.

Hi, we need the possibility to add an LLM.txt and schema markup to sites on the basic plan. This is a standard feature of a modern website and shouldn't be gated behind the Pro plan or higher. Moreover, it's one of the actually useful use cases for an AI that still can't be trusted to make design decisions. I can generate Schema Markup in seconds on Webflow for all pages, and generate an llm.txt file easily also; I should be able to do so here, on any level. Thank you.