Released October 30, 2026.
This release is about the people who read your docs through an AI tool. Sign-in codes for your Docs MCP server now come from your own domain instead of from Leed, and AI agents connected to your docs can now read full pages, not just search results. It also cleans up the access lists on page types and labels.
New features
Docs sign-in codes come from your own domain
When a reader connects an AI tool to your Docs MCP server and signs in, the one-time code they receive now comes from no-reply@ your site’s domain, with your site’s name as the sender. The subject reads “Your site name docs sign-in code”, and the email no longer mentions Leed anywhere except the small “Powered by Leed” line at the bottom. Readers see an email from the company whose docs they are reading, so it is less likely to be ignored or caught as spam.
Your domain has to be set up to send email for this to work. If it isn’t yet, nobody is locked out: the same code is sent once more from Leed’s address, still with your site’s name in the subject and body. A temporary delivery problem is retried as before and never switches the sender. The email you get when you sign in to the Leed CMS itself is unchanged.
Site names with commas, quotes or accented letters, such as “Acme, Inc.” or “Café Labs”, now show correctly as the sender name. That also applies to your campaign emails.
Learn more: Reader Sign-In Flow, How Email Works in Leed, Bounces and Deliverability.
Fixes
AI agents can read your full docs pages
On some sites, an AI agent connected to the Docs MCP server could find pages through search but got a “not found” error when it opened any of them. The site wasn’t publishing the machine-readable copy of each documentation page that the agent reads, because a setting that was never shown in the editor had been left off.
Documentation and API sections now always publish those copies, and that setting can no longer be turned off. As a result, your documentation and API pages are also listed in your site’s llms.txt and its feeds. Existing sites pick this up the next time their settings are published and the site rebuilds.
Learn more: What Your Visitors Get, Feeds, Sitemaps and robots.txt, Page Type Reference.
Access lists no longer show “undefined undefined”
If you publish your docs from a CI pipeline, the access lists on your page types and labels showed an entry named “undefined undefined” for that pipeline, and removing it didn’t stick: it came back on refresh. Those entries are now left out of the list. CI publishing is unchanged and is still managed with its trust. An entry for someone who is no longer a member of your company now shows their user ID and can be removed.