18th May – 2026 release notes: Chat uploads everywhere, project folders, and a cleaner Compliance Agent

This week’s release focuses on three major improvements.

First, you can now add files and images to all remaining chats in Magnity.
Second, we are releasing folders for projects.
Third, we have significantly reduced the initial information load in the Compliance Agent to make review easier and faster.

📎 Add files and images to any chat

Last week we released the ability to add files to chat in the Rules Agent and Persona Agent using the small + icon on the left side of the chat.

That release went well, and we are now rolling out the same capability to all remaining chats in:

• Email
• Webpage
• SoMe
• Content Agent

Compliance Agent will follow as well, but we still need to implement a few agentic pieces there first.

You can now upload:

• PPT
• DOCX
• TXT
• PDF
• Images

This means you can feed a chat with much richer context.

Examples:
• Add an image of a webpage and ask Magnity to extract the content and give you a conversion optimization analysis
• Add campaign plans as screenshots or images and ask Magnity to build from them
• Upload visual concepts you want to discuss or improve
• Use uploaded images as the starting point for image editing or re-use

You can also choose Add content manually, which opens a field where you can simply paste in text as a manual source.

This makes chat much more flexible and much better suited for real work, where your input is rarely just one tile or one page.

📁 Folders for projects

We are also releasing folders for projects.

While this sounds simple, it is a meaningful improvement to how you organize work in Magnity.

You can now:

• Create folders
• Delete folders
• Create nested folders
• Drag and drop projects into folders
• Move a project into a folder using the ... menu
• Select what folder to save a project in directly from one of the agents

This should make it much easier to keep your projects organized and manageable over time.

✅ Compliance Agent: reduced information load and easier review

We have made a significant update to the Compliance Agent interface to reduce the amount of information shown at once and make findings easier to work with.

What has changed

Removed left-side finding details panel
We have removed the entire left tab that previously showed finding details.
Instead, details now live directly under each finding card in the Analysis tab and can be expanded by clicking See details.

Replaced explanatory text under category headlines
We removed the explanatory text below the findings section headlines and replaced it with pills showing the severity level distribution for the category findings.

Redesigned finding cards
We have redesigned how each finding card works.

First, we now show the suggestion by default instead of the description.

Suggestions contain two fields:

• Replace this
• With this

“Replace this” is tinted red/coral, and “With this” is tinted green. This makes the key action much easier to understand immediately.

Second, we removed the subcategories called Resolved and Dismissed as separate movement states in the interface. Previously, cards moved around when state changed, which created confusion. Now cards stay in place regardless of whether they are:

• Finding
• Fixed
• Resolved
• Dismissed

You can still filter by these states if needed.

Third, each card now has a very clear color and matching pill/tag based on state:

• Yellow = Finding
• Purple = Fixed
• Green = Resolved
• Coral / Red = Dismissed

The matching text in the raw source is highlighted accordingly, making it much easier to understand the status of each issue.

⏭ Coming very soon to Compliance Agent

We are already working on the next set of additions, likely for early next week.

Comments on findings

Each finding will be commentable by users, so you can add extra information for reviewers or coworkers.

Add finding manually

You will be able to highlight text and add a finding manually.
When text is highlighted, a small hover button will appear and open a new finding card you can edit.

Feedback when dismissing

When dismissing a finding, you will be asked to give feedback. This will not be mandatory, but it will help teach the system where it fails and support future proactive rule updates.

Select document type

We will add the ability to define a document type, such as:

• Email
• Article
• Press release
• Webpage

This will become available as a merge field in rules, so you can write rules specific to document type.
For example, for email analysis you may not want to review the header and footer.

We do not have a release date yet, but we hope as soon as next week.

🛠 Other fixes and updates

Word upload fix in Compliance Agent

We fixed an issue where Word documents with many nested tables were not read correctly.
They are now handled properly in the Compliance Agent.

Flow Builder improvement

We adjusted the Flow Builder so that subject line and preheader are now built in sequence.

This means the preheader now knows what the subject line is before it is created, just like in the ordinary email builder.

Email exporter safeguard

We adjusted the email exporter so unsaved email projects can no longer be exported.

This prevents several potential file-related issues.

🤖 Agentic backend preparations across all builders

We have now made preparations so that all builders have an agentic backend.

For the more technical among you, this means we can begin introducing inter-agent connectivity.

Examples of what this could enable:
• Invoke the Rules Agent from the Email Agent
• Invoke the upcoming SEO/GEO Agent, or Findability Agent, from the Webpage Agent
• Invoke the Compliance Agent from the Content Agent

This opens up many possible concepts and workflows.

It is also an important path toward a stronger review and collaboration setup around content creation. There is still work to do before this is fully connected, but the groundwork is now in place.

🧠 Learning from first draft and final export states

We are also beginning to store emails, webpages, and SoMe posts in both:

• First draft mode
• Final export mode

Over time, this can be fed into the Rules Agent.

The purpose is to move toward self-correcting agentic networks that learn from your interactions and suggest rule changes as a consequence.

This is not finished yet, but development is underway.

🔮 Looking ahead

We have more things underway that we believe will continue pushing the marketing space forward, and we will share more as we get closer.

The long-term goal remains the same:

To build an Agentic OS for B2B marketing that behaves like an autonomous marketing agent.

These releases are all steps in that direction.

Questions or feedback?

Our support team is here to help:
✉️ ulsa@magnity.ai | 📞 +45 29 35 25 78

Thank you for using Magnity.
— The Magnity Development Team