Connectors
Webflow
The tool reads a Webflow site through a site API token and copies it to an Insites v6 instance: every collection and every item in it, every image and file, every page the site publishes with its design, its collection pages and lists, its redirects and sitemap, and what people sent through its forms. It only reads. Nothing is ever written back to Webflow. Migrate a Webflow site walks through a whole migration.
Make a site API token
Only an administrator of the site can make one.
- In Webflow, open the site you are migrating, then its Site settings (the gear icon).
- Open Apps & integrations and scroll to API access.
- Click Generate API token and name it
Insites Migration Tool. - Set the permissions in the table below. Leave everything else at No access.
- Click Generate token and copy it: Webflow shows it once. Paste it into the tool with the address visitors use for the site.
A site token reads only the site it was made in, so make it from inside the settings of the site you are migrating.
Permissions
| Permission | Set it to | What it lets the tool read |
|---|---|---|
| CMS | Read-only, required | Every collection and every item in it, drafts and archived ones included. |
| Pages | Read-only, required | Which pages the site has, and which of them show a collection's items. |
| Assets | Read-only, required | Every image and file uploaded to the site. |
| Sites | Read-only, recommended | The site's name, its domains and its languages, and on an Enterprise plan its redirects. Without it the tool reads the site's id from its home page instead. |
| Forms | Read-only, recommended | The site's forms and everything people have sent through them. |
| Ecommerce | Read-only, if the site has a shop | Products and orders, counted so the inventory names them. They are not copied in this version. |
When you connect, the tool checks each permission with one small request, and the connect page lists every one of them: Can read, Missing for a required one the token lacks, or Not given for a recommended one, with what that leaves out. Without a required one it does not start. Without a recommended one it starts, and says what it cannot count.
It does not ask for Custom code or Authorized user, because a site token cannot use them, or for Users, because Webflow switched off its User Accounts API on January 29, 2026. Comments, Branches, Site activity and Workspace are not site content. Leaving them all at No access means the token can do as little as possible if it ever leaks.
What is copied
| In Webflow | On Insites |
|---|---|
| Each collection | A table in the Data module (Databases in the admin), named after the collection's slug and labeled with its name. Its fields keep their names, help text and groups, with a group called From the old site (reference only) for what Webflow kept about each item: its source id (the Webflow id), Draft, Archived and when it was last published. Changing those fields does not change the site, except Draft and Archived in a collection with no item pages. The Published address is not in that group: it sits under Slug. |
| Each item | A row, with the dates it was created and last changed. Drafts and archived items are copied too, marked, so nothing is lost. Each item's Published address decides whether it is on the site: an item is shown, listed and in /sitemap.xml while it has one. It sits right under Slug in the admin, with help that says to fill it in. The migration fills it with the slug of every item published on Webflow, and leaves it empty for drafts and archived items. Draft and Archived are kept for reference and hide nothing, except in a collection with no item pages: see Each collection list below. |
| Images and files in image and file fields | Copied into the row's upload column, public, as they were on the live site. A file name with brackets, such as "photo (1).jpg", keeps its whole name and its extension. A name Webflow saved with an encoded percent sign ("%2520", for "%20") is copied under its readable name, so it opens. An image's file name is never replaced with a field's value, even when an item's slug is the same word. |
| Images inside rich text | Copied to Assets, in the folder ins_assets/site/cms. The text names the copy instead of Webflow's file host. |
| Multi-image fields | A multi-image field in the Data module, shown in the admin with an image uploader holding the item's images, in Webflow's order. Each image is copied from Webflow into its own attachment, the way the admin stores an image uploaded there, so editors add, remove and download images as they would in any Insites database. The collection page shows them in the field's order. |
| Links in rich text to the site's own pages | Made relative, /about rather than https://www.example.com/about, so they lead to the same page on the new domain. With Sites access, a link to any of the site's other domains is made relative too. |
| Canonical and language links | Kept as full addresses, https://www.example.com/about, as Webflow writes them. A page's canonical link and its hreflang alternates name the site's real domain, which stays the same after the site goes live. |
| The site's asset library | Every file, copied and checked. Images, documents such as PDFs, video and sound go to Assets in the new instance's admin, in the folder ins_assets/site (stored at modules/ins_assets/public/assets/site/), where editors find, replace and download them. Stylesheets, scripts, fonts and other files the design uses go to assets/site/. The Assets screen shows this as a second, top-level site folder that looks empty, because the screen does not list those files. Nothing is missing: the images are in ins_assets/site. Do not delete or rename the top-level folder, or the site loses its design. |
| Each published page | An Insites page at the same address, read from the live site as a visitor gets it, with its title, description and sharing tags. The pages share one layout, site, which keeps what Webflow's interactions need, so animations and menus work as before. Each page answers only at its own address, as on Webflow: an address below it, such as /pricing/old-plan under /pricing, answers 404 with the site's not-found page, and a redirect set for that address in Settings, Redirects still works. Without this an Insites page answers every address below its own. |
| The design | The site's stylesheet, scripts and fonts, copied to assets/site/, and the images it shows and their smaller versions for phones, copied to Assets in ins_assets/site. Each page names them with asset_url, so nothing loads from Webflow any more, and the stylesheet names its images by paths relative to its own copy. |
| Each collection page | An Insites CMS collection at the same addresses, such as /our-work/clarins, which you manage in CMS > Collections. Its details layout shows each item: the tool learns it by lining up the item pages Webflow publishes, where everything they share is the layout and each place they differ has to be one of the item's fields. Its list layout is the page Webflow had at the collection's own address, such as /our-work; if Webflow had none, that address answers 404, as on Webflow. An item with no Published address, or a made-up address, answers 404 too. /sitemap.xml lists every item with a Published address, unless Webflow's own sitemap left the collection's item pages out: then the collection stays out of /sitemap.xml too, and its items are still on the site. A collection of more than 1,000 items is named in the plan: the Insites sitemap lists only the 1,000 most recently changed. The collection's page is not in the instance's site search, so a blog index that becomes a collection's list page is not in it either. |
| Each collection list | A list that reads its table, so an item an editor adds or changes shows up. The tool works out the list's filter, order and limit from what the live pages show, and its pages too, with Webflow's next and previous links. A paginated list that starts after its first items (Webflow's Start from) shows as many items a page as Webflow does. An entry whose text field is empty, which Webflow marks w-dyn-bind-empty, is read the same way: the mark follows the field, so filling the field in the admin shows it. A list that an earlier version kept as fixed text is rebuilt when you run the move again. Which items a list can show depends on the collection: a collection with item pages, such as /our-work/clarins, lists only items with a Published address, as its item pages show only those. A collection with no item pages, such as testimonials or logos, has no address to fill in, so its lists show every item that is not a draft and not archived. For such a collection, Draft in the old site and Archived in the old site hide an item from its lists while either is on, and the admin's help on those fields says so. A collection counts as having no item pages when Webflow publishes no template page for it, or when every item page the tool reads answers 404 on the live site. The tool tells which item each entry shows by the link to its page, a file only that item has, its name, or a value only it has, in that order, and accepts a list only when every entry is one item and no item shows twice. A list inside another list's entry, such as a list of months each showing the articles its multi-reference field names, reads those items for each entry, in the order the field holds them, so an article linked to a month in the admin shows under it. A list of entries whose reference, file or image field is set (Webflow's "is set" filter) leaves out the entries where it is empty. A multi-reference field is picked by name in the admin, from the collection it points at. A list it cannot work out keeps showing what the live site showed on the day it was read, and the plan names it with why; the other lists on the same page still read their tables. |
| What people sent through a form | A table for each form, one column per field and one row per submission, with when it was sent. A later sync adds what has been sent since. Webflow keeps submissions made through every copy of a form, including one that was deleted, replaced by a copy, or is no longer on a page. Those load into the table of the form with the same name, and each one's uploaded file, such as a resume, is copied into the file column with the others. That holds for a form no page draws any more, too: its table still gets a file column, and each resume is copied rather than kept as Webflow's address, which opens only with a Webflow login. |
| Each form on a copied page | Rebuilt in the Insites form builder. See Forms below. |
| The sitemap and robots.txt | The sitemap is the one Insites serves at /sitemap.xml. It lists every page that was copied, except the not-found page and any page Webflow's own sitemap leaves out. A collection's items are listed through their collection. robots.txt keeps the site's own rules and points to /sitemap.xml. |
| Scripts from other services | Kept as they are, such as analytics, a booking widget or a forms service. An Adobe Fonts web project serves any site with its embed code, so its fonts keep working. The check asks each kit as the new instance, and the To do list has a job only if the kit refuses it. |
| Redirects | Each becomes a page at the old address that redirects to the new one. The tool finds them itself: see Redirects below. |
Forms
Each form on a copied page is built the way the Insites form builder builds one, so you edit it in Forms in the new instance's admin like a form made there:
- Its database table comes first: the same table its past submissions go into. A form nobody has sent yet gets an empty one.
- Then the form, on that table, with one field for each field of the Webflow form.
- Then its two emails: a workflow email to the site's owner for each submission, and an autoresponder to the person who sent it.
The page keeps the form as Webflow published it: the same fields, labels, button and styling, and the same messages after sending and when sending fails. It now sends to the form on Insites. Webflow's own script leaves it alone, because the form has an address of its own to send to.
Webflow's API does not say who its form emails go to, so the Forms step of Transform asks who each form's workflow email goes to. You can give more than one address, separated by commas. A form you give no address to saves every submission and emails nobody, and the plan says so. The autoresponder is made switched off, so nothing is sent until you have read it and turned it on in the form builder. Replies to it go to the first address you give, or to the address the old site links to until you give one. Each enquiry also adds a CRM contact for the email address it was sent with, unless you untick Adds a CRM contact for the form. Each enquiry adds a CRM contact. Both emails are sent from the new instance's own sender, the way the form builder sends them, which needs the CRM module at 6.1.2 or later: if it is older, the plan is blocked until you update the CRM in Insites Console.
A form on several pages keeps the look it has on each page. If the site styles its contact form one way on the contact page and another way in a band on every other page, each page keeps its own, and both send to the same form on Insites. A page that shows the same form twice in two styles, such as in a hero and in the footer, keeps each copy's own style. Two different forms that share a name, such as Webflow's default "Email Form", stay two forms on every page that shows them, including a collection's item pages, and each keeps its own fields.
A form that used Webflow's reCAPTCHA uses the platform's hCaptcha in its place, because Webflow's reCAPTCHA key works only on Webflow. Send the form once on the new site to see the check. Once every form uses hCaptcha, Webflow's reCAPTCHA script is taken out of the copied pages, because nothing calls it any more.
A file upload field, such as a resume on a careers form, stays on the form with Webflow's look and messages, and uploads the file to an upload column of the form's table on Insites, kept private so only the site's administrators can open it. It keeps Webflow's size limit and file types. Webflow's own upload script no longer handles the field, so a file is never sent to Webflow. Each past submission's file is copied into the same column with the site's token. This holds for a form no page draws any more, such as an old Careers form: its files are copied as files too, not kept as Webflow's address (which opens only with a Webflow login), and the SOURCE check agrees with what was copied. The file's address is taken from whichever part of Webflow's value holds one. When Webflow sends a file as just its id and name, with no address, the tool uses Webflow's own address for it (webflow.com/files/<site id>/formUploads/<name>) and copies it the same way. Webflow can keep these files private: with Restrict uploaded file access on in the site's Site settings, under Apps & Integrations, a file answers only to someone signed in to Webflow, and Webflow gives no way to read it with the site's API token. Such a file is not copied, and the To do job Find the files the old site would not give says so. These files are often sensitive, such as ID or eligibility documents, so take care. Only turning the setting off needs the client's OK: while it is off, anyone with a file's link can open it. Downloading each file from Webflow while you are signed in and uploading it to its row in the new instance's admin exposes nothing, so it needs no OK, and it is the safest way. If you turn the setting off instead, get the client's OK first and keep the time short: turn it off, press Copy again on the files card on the Verify step, then turn it back on straight away. Webflow may keep files uploaded while the setting was on private whatever the setting says, so some may still need to be downloaded and uploaded yourself. A past submission whose value names no file is left empty, and the plan says in one line how many of the form's rows that is, with an example. The Upload File button opens the file picker on a click, Enter or Space, as on the live site: Webflow's script is kept off the upload block so it cannot cancel the click.
A Webflow form that has no submissions and is on no page is not rebuilt, and no database is made for it. It is listed under Left behind in the plan, the report and Verify, with the reason no submissions, not on any page. A form with submissions keeps its database even when no page shows it, and a form on a page is rebuilt as above.
When you leave a form behind on Transform, Code files left behind, a page, shared section or page frame that still draws it would show "Could not find Form with name" on the new site. The plan takes the form out of those files for you and lists each one under the changes it makes, so you do not meet the error on Verify. The rest of the page is copied as it is.
Every Webflow template ships a /style-guide page of sample content. When nothing on the site links to it, the plan says so under its warnings and offers to leave it behind. It is copied by default, because the page may have been used. To drop it, untick it on Transform, Pages and code and give a reason. A page left behind can be copied again only by a later sync while the old site still exists.
When you move a site again and the new site already holds a form from the earlier move, the form keeps the name it has there, numbered or not, and so do its table and every page that shows it. The files of its two emails follow the earlier name too, so the check no longer reports the autoresponder as missing.
If an earlier migration into the same instance made the file field a text column of file addresses, the next migration replaces it with an upload column and copies each past submission's file into it. It does this only when every text the column holds is a file address the migration copies. If any row holds other text, the plan stops, names the rows, and changes nothing: clear or move that text on the new site, or leave the field out.
Webflow draws an image with no picture chosen using its own placeholder, an address under cdn.prod.website-files.com/plugins/ that Webflow refuses to everyone, so it shows as a broken image on the old site too. There is nothing to copy, so the page draws an empty image in its place, whether or not you choose to replace Webflow's file addresses, and the plan lists it under Left behind.
A form from another service, such as JotForm, HubSpot, Typeform or Google Forms, is that service's form embedded in the page, not a Webflow form: Webflow's API does not list it. The page keeps it as it is, so it still sends to that service, which keeps what people send and sends its emails. Nothing it sends is recorded in Insites. The plan lists each one under Forms from another service, with the pages it is on, and the To do list asks you to decide where it should send.
To keep its submissions in Insites instead, open Transform, Forms and press Rebuild it as an Insites form beside it. The tool reads the form's own public page, and AI proposes the Insites form: its name, each field with its kind, whether it is required and its choices, and the thank-you message. Check each field, change anything, say who the workflow email goes to, and press Build this form in place of the JotForm form (or the service it is from). Then save your Transform choices. The plan then builds it in the Insites form builder: a database for what people send first, then the form on it, checked with hCaptcha, with a workflow email from the new instance's sender and an autoresponder made switched off. Every page that carried the embed shows the Insites form in its place. Before you confirm, the proposal lists what the rebuilt form will not do that the JotForm form does, such as rules that show or hide questions, pages, hidden fields or calculated fields, and the To do list says how to do each one again if you build it anyway. Until you confirm, nothing changes. Keep the JotForm form instead leaves the embed as it is, before or after you confirm.
- A date asked as Day, Month and Year is one date field. A question whose only boxes are a day, a month and a year is proposed as one date field named by the question, required if any part was, with or without AI. A question asked as checkboxes keeps every choice.
- File uploads are not built. A file field in the form is listed as left out. Add a file field in the form builder after the load: it uploads through the instance's own upload settings.
- Past submissions stay with the service. Export them there if you want them.
- HubSpot forms cannot be read from outside. HubSpot draws its form from your HubSpot account with its own script, so there is no public page of the form to read, and the tool says so. Build that form in the form builder yourself.
- A form in a shared section, such as a footer on every page, is not moved there: the plan says so.
- The AI is not asked to make a form up: a page where the tool finds no fields gets no proposal. With the AI pass off, or once the migration has reached its AI limit, the proposal is the fields as the tool read them, without AI.
Older submissions Webflow's API does not give
Webflow can list more submissions for a form than its API gives. On one site the dashboard listed 75 for a form and the API gave the newest 47, from July 2023 on: the older ones are in Webflow's dashboard and its CSV export, but not its API. Transform, Forms shows each form's count. If Webflow lists more:
- In Webflow, open Site settings, then Forms, and export the form's submissions as a CSV.
- In the tool, open Transform, Forms, and add the file under Older submissions from Webflow, beside the form's table. There is one field for each form that saves to a database, on a new run and on a Run again, and it is there before the dry run.
- Save your Transform choices. The plan adds each line as a record, with the date it was sent. The submissions already copied from the API are matched by what they say and when they were sent, so none is added twice, and giving the same file again adds nothing. The match ignores what nobody typed: line breaks written differently, extra spaces, the case of an email address and the spacing in a phone number. Each submission already there matches one line of the file at most, so two real submissions that say the same thing a minute apart stay two. If a line is added although it differs from a submission already there in only one field, the plan names the line and the field, so you can check it.
A column in the file that matches no field of the form, such as Webflow's own id, is named and left out.
A collection whose name is about forms, such as Contact Forms, holds Webflow content, never what people sent. Its table is labeled (Webflow collection, no submissions) in the admin, so nobody looks for submissions there. The same goes for a collection whose items hold the ids of the forms embedded on the site.
A new instance on the Insites staging stack does not deliver form emails to the people you name: staging sends every email to the Insites team's test inbox instead. There, check that each submission is in the form's database and that the page shows its thank-you message, and test that the email arrives once the site is on a production instance.
Redirects
Webflow lists a site's redirects through its API only on an Enterprise plan. On that plan the tool reads them from Webflow. On any other plan it finds them from the outside, with nothing for you to export:
- It gathers the site's old addresses: every page address the Internet Archive has seen on each of the site's domains, and every link on the site's pages and in its CMS that leads to a page the site no longer publishes.
- It asks the live site about each one, without following the answer, and keeps each one that sends a visitor to another page.
- A chain is shortened, so a visitor takes one hop to the page at the end of it.
The plan says how many old addresses were asked about and how many redirect. Finding none this way does not mean the site has none. Webflow does not share its redirects with the tool, so the Redirects step says the result is not confirmed, in amber, and the To do list carries a job to confirm it. The job is done when you add Webflow's export, or tick This site has no redirects and say why (the reason goes in the report). It never stops the load. When you add an export, the step shows the count from your file and keeps the live-site check as supporting information. A redirect that no archive saw and no link names cannot be found this way. To be sure of every one, export the list from Site settings, Publishing, 301 redirects, and add the CSV on the Redirects step: the tool reads it as Webflow writes it, the old path first and the new one second, and a list you give wins over what the tool found for the same address. An old address with a query string becomes a page at its path, because a page answers its address whatever the query string.
Redirects set in Settings, Redirects work only when the new site's not-found page hands an unmatched address to them. A 404 page copied from the old site, such as Webflow's, lacks that line, and every redirect would answer 404. The load adds the line to a copied 404 page that does not have it, so the redirects answer. A sync adds it too: a 404 page written by an earlier move without the line is not taken for a change your team made on the new site, so the next sync puts the line in. A 404 page your team did change on the new site is still kept as it is, with only the line added at the top.
Webflow's redirect syntax is read as Webflow means it. A % in front of a character in the old path escapes it, so /news%-and%-social is the address /news-and-social. A (.*) wildcard that is a whole piece of the old path, such as /news%-and%-social/(.*), becomes one page that catches every address under /news-and-social/, and %1 in the new path is what it caught, so /news/%1 sends /news-and-social/2024/launch to /news/2024/launch. Anything else Webflow reads as a pattern, such as alternatives like (news|press), a character class like ([0-9]+), a wildcard inside a piece of the address like /blog-(.*), or a wildcard that would catch every address on the site, has no Insites page address that answers the same addresses. Each one is not moved, and the Redirects step and the plan name it with the reason.
Each kind of field
| Webflow field | Insites column |
|---|---|
| Plain text | Text, on one line or several as in Webflow |
| Rich text | Text shown as HTML |
| Image | Upload, with the image control |
| File | Upload, with the file control |
| Multi-image | A list, with the File (Multiple) control: the image uploader, holding each image as an attachment |
| Link, Video link | Text, with the link control |
| Email, Phone, Color | Text |
| Number | A whole number or a decimal, as the field's settings say |
| Date/Time | Date and time |
| Switch | Yes or no |
| Option | Text, with a dropdown of the option names |
| Reference | Points at the other collection's table and holds that row's id, with a picker in the admin |
| Multi-reference | A list of the other rows' ids |
| Item id, draft, archived, last published | source_id, is_draft, is_archived and source_published_at. Draft and archived are kept for reference and hide nothing |
| The address an item answers at | published_slug, labeled Published address, right under Slug: the item is on the site while it is filled in. The migration fills it with the slug of each item published on Webflow, and leaves it empty for drafts and archived items. Every link to an item, every list and the item's own page read this one field, never Slug, so a link always leads to the item's page |
A kind of field this list does not name is copied as text, and the plan says which.
Ids, references and later syncs
Webflow names each item with 24 letters and numbers, and Insites numbers its rows. The plan gives each new item the next number above every id the destination has used, and keeps the Webflow id in source_id. A final sync finds each row again by its Webflow id and updates it, so nothing is added twice. The destination numbers the rows an editor adds later above these, so they never clash.
A reference follows the item it names to that item's new number. A reference to an item that is not on the site any more is cleared, and the plan counts it.
Reference fields and blog lists on a collection page
A collection page can print the item a reference field names: a blog post's author, its category, and the link to that item's own page. The page reads the row the reference points at and prints its fields, so changing the author's name in Data changes it on every post. A list on a collection page that shows only the items naming that page's own item, such as a category page listing its posts, is rebuilt as a list that keeps exactly those items, by single or multi-reference. The plan describes each such list, for example "posts items, the ones whose category names this page's item, newest first, all of them". Collection pages are rebuilt while the source is read, so a site read before version 0.8.12 needs its collection pages rebuilt for its blog to be rebuilt this way. Press Rebuild the collection pages, as described below.
How a collection page is checked before it is copied
Before a rebuilt collection page is used, the tool fills it in with each item's own values and compares the words with that item's live page. If one item's page says anything different, the template is not used. The same goes for a list whose entries no filter and order of Webflow's can give: it keeps showing what it showed when the page was read, and the plan says so.
Rebuild the collection pages of a site already read
For a Webflow migration that has been read and not loaded yet, Pages and code on Transform has a Rebuild the collection pages button. It reads the site's pages again and rebuilds its collection pages and lists with this version of the tool, so a site read by an earlier version gains what this one builds, such as blog post and list pages. Every record, file, form submission and redirect from the earlier read is kept as it was, so it takes minutes, not a whole new read. The step says Rebuilding the collection pages... at once, and the activity log says Rebuilt the collection pages with how many were rebuilt. The plan is built again, so you type its code again before the load. Operators can press it. Roles.
When the item pages alone cannot explain the template
The usual reason is markup only some items show, such as a block with Webflow's conditional visibility. When the AI pass is switched on for the tool, it asks Claude to propose the template for the places that differ, in the tool's own form: each place prints one of the collection's fields, or shows some markup only when a field is set. The proposal is used only when every item page renders word for word, with the same elements and attributes, and the plan says it was proposed by AI and verified. The activity log says which model answered and how many tokens it took.
What is sent: the places the item pages differ, as they read on two or three item pages, the collection's field names, and the values of those items. Keys, codes and email addresses are taken out first. Whole pages are never sent. At most one request goes for each collection, and a second only when the first answer cannot be used at all.
When the proposal does not verify, or the AI pass is off, you choose on Transform, under Pages and code, for each collection: use the proposed template (the difference it has is written into the plan and the report), copy each item page as a static page (the site reads the same, the rows still move, and a row added later has no page until a layout is written), or leave it for you to rebuild. The plan names each collection page left for you with its reason. Until you choose, its item pages are copied as static pages. When Webflow has no page at the collection's own address, such as /blog-category, a page there answers 404, as on Webflow; it also takes the place of a CMS collection page an earlier move of the same site left at that address, so /sitemap.xml does not list the items twice.
Webflow does not say which filter or order a list uses, so the tool picks the likeliest one that gives exactly what every page shows: newest created first, then the smallest offset, then the simplest filter. The plan lists what it picked for each list, so you can check it.
What is not copied yet
What moves from Webflow lists every Webflow feature, CMS fields, collection list settings, images, interactions, forms and site settings, with what moves and what to do about what does not.
- Draft pages, which are not published, and password-protected pages, whose content is not public. Both are counted, with the reason.
- Ecommerce: products, SKUs and orders. The site's pages, collection templates and 404 page are copied. The currency template in each page head is kept as plain text, so these pages are not held back as code the new site would refuse.
- Other languages. Only the site's main language is copied. The plan names each other language left behind.
- Site members, which Webflow no longer lets any tool read.
How it is checked
The same checks as any migration, and these that matter most for Webflow:
- SCHEMA: every table and column is set up as planned.
- REFERENCES: every reference points at a row that is there.
- SOURCE: the tool reads Webflow again at the check, and every row matches the site as it is at that moment, each found by its Webflow id.
- CONTENT: every page says the same on both sites, word for word, every item page of every collection included. A page Webflow was too busy to serve counts as not compared, and fails.
- ADDRESSES: nothing a page or its stylesheet loads still comes from Webflow, and every file a page loads answers. A file it could not read to check is named, and does not pass.
- REDIRECTS: every old address, found or given, answers with a redirect to its page.
SITE and SCREENS compare each page's status, title and look on both sites, as they do for any migration. The checks lists them all.
How fast it reads
Webflow limits how many requests a token makes in a minute, and the limit depends on the site's plan. The tool reads the limit from each of Webflow's answers and stays under it, one request at a time. If Webflow asks it to slow down, it waits as long as Webflow says, and if the wait would run past the step, the run pauses and carries on by itself. A site with a few hundred items is read in under a minute.
If the tool says
| Message | What to do |
|---|---|
| … is not a published Webflow site. | Use the address visitors use, on a site that is published. The tool reads the site's Webflow id from its home page. |
| The tool could not reach … | Check the address, and that the site is published. |
| Webflow did not accept this token. | Make a new site API token in the site's settings and paste it again. |
| The token cannot read CMS, which the migration needs. | Make a new token with the permissions it names set to Read-only, and paste it. |
| Webflow has no such site for this token. | The token was made for another site. Make it inside the settings of the site you are migrating. |
| Webflow is not answering, or is limiting how fast the tool reads. | Nothing: the run waits and carries on. If it keeps happening, try again later. |
When the migration is done
The To do list in Launch lists the jobs this site needs, and these come with a Webflow site:
- The site's other domains. A Webflow site often has several domains, each sending visitors to the main one. Each needs pointing at Insites too, and the checklist names every one.
- Adobe Fonts, only if a kit refuses the new address. The check asks Adobe Fonts for each kit as the old site and as the new instance. Web projects serve both, and then there is no job. An older kit that lists the domains it serves can refuse the new address: the job then names the kit and asks you to add the new address and the site's domains to its list. If Adobe Fonts could not be reached, the job says so and asks you to look at a page's fonts.
- Forms. Open each form in the form builder: check who its workflow email goes to, read its autoresponder and switch it on, and send a test from the copied page.
- Collections. Show the people who edit the site where each collection is now: its items in Data (Databases in the admin), its list and item pages in CMS > Collections. To publish an item, set its Published address to its slug. To take an item off the site, empty its Published address. Draft and Archived are kept for reference and hide nothing. An item added in the admin is not on the site until its Published address is filled in: the tool does not fill it in from Slug, because an empty Published address is how an item is kept off the site. When you duplicate an item, give the copy its own Published address, or the copy takes the original's page.
- Pages, the layout and the stylesheet. Show the people who edit the site where the code is now, because a collection's page is not where they look first:
- A page with no collection, such as
/about, is in CMS > Pages. - A collection's page, such as
/portfolioand each/portfolio/<item>address, is not in CMS > Pages. The admin lists a page that reads a collection only under CMS > Collections. Open the collection there: the List Layout tab holds the code of the page at the collection's own address, and the Details Layout tab holds the code every item page shares. - The layout every page shares is
site, in CMS > Layouts. - The site's stylesheets, scripts and fonts are in
assets/site/. The Assets screen does not list them, so edit them in CloudShell, or change them in the code and deploy.
- A page with no collection, such as
- Search engines. Submit
/sitemap.xmlin Google Search Console once DNS has moved. - Keep the Webflow site published, with its hosting plan, until the new one has run cleanly for two weeks: pointing DNS back is the fastest way out. Then end the plan.
The checklist says all of this for a Webflow site: it never tells you to delete an instance in Insites Console, which is the last job when the source is an Insites instance.
Then revoke the token in Site settings, Apps & integrations, API access. The tool removes its own copy when you close the migration: how long a key is kept. More on keys and data.