October 15, 2026 / 🌶️ Create Custom Integrations via AI fast (preview), Rich text changes history in your Inbox, MCP Server can execute code

[mdubakov]

Create Custom Integrations via AI fast (preview)

Need Fibery to sync with a tool we don’t integrate with yet? Don’t wait for us (most likely we will never do it). Describe what you need, and AI builds the connector, tests it and keeps your data syncing. This is a preview release, so you may spot some rough edges here and there, but we tested it extensively for many external tools.

To start, click + in the sidebar → Space → Connect via AI and ask:

• “Sync deals and companies from Attio”
• “Sync my books from open library"
• “Sync tickets from our internal API at api.example.com”

Some important things for you to know:

• Hosting is free during the preview (~2-4 weeks), up to 4 hours of compute per day per workspace. After that, we’ll most likely charge for compute time: the more your connectors run, the more you pay. What we see during the preview will shape the fair price. Note that Custom Integrations are running on Vercel now
• Connectors support OAuth and webhooks (changes arrive as they happen, not at the next sync). It also can create Custom Actions that you can use in Fibery Automations to update data in external system
• Building and editing connectors uses AI credits, like any AI chat in Fibery. Lighter models such as GPT-6 Luna or Claude Haiku usually do the job, so try them first
• Current limits: HTTP APIs only, data flows into Fibery only, and there’s no MCP tool for creating connectors yet

There’s no video walkthrough this time, on purpose. We want to find rough edges with you, brave early adopters! Build the integration you’ve always wanted and tell us how it went in the community. Good luck!

Rich text changes history in your Inbox and entity history

Finally changes in rich text are properly rendered in entity history with diffs. When you click on a change, you will be redirected to a proper place in the document.

Moreover, now you will receive notifications about changes in rich text fields. Notification is sent when a new version is created in the rich text (about 10 mins of inactivity after the first change).

MCP Server can execute code now

We’ve added execute_code tool to Fibery MCP. AI agents can now write a small script that calls other Fibery MCP tools. It’s useful when an agent needs to chain tool calls, or to process output of the tool with JS right away instead of looking at the raw output. For example, it can aggregate data more efficiently during research and set up multiple parts of workspace in parallel.

In a nutshell, it saves tokens now. In some situations a lot of tokens.

Type filters right into search (experimental)

You can now narrow search down without reaching for the mouse. Type in: to pick a database, then your search text. Or add the filter afterwards, once you realize the entire workspace was a slightly ambitious place to look. Repeat in: to look in several databases at once.

Suggestions appear as you type: arrows to choose, Enter or Tab to confirm, ? to list every filter. You can edit anywhere in the query without disturbing the rest.

A few more filters to play with: created:, by:, people:, and is:Unfinished, plus space: when you’re hunting for views (in:Boards space:Marketing). Prefer clicking? Filter changes made through the UI now appear in the query too. A small syntax lesson nobody signed up for.

in: also works in # and [[ mentions inside documents.

It’s an experiment for now, so turn on Search operators in Settings → Experimental Lab and tell us which filter we forgot.

Text table UX improvements

Right-clicking in a table now opens your browser’s menu instead of Fibery’s, so you can copy text, check spelling, and use the usual options.

Table actions now live where they belong: cells, rows, and columns each get their own menu.

Performance improvements

Faster Views (and everything around them). Workspaces with lots of Views should now feel more responsive whenever you setup filters, sorting and columns, rename or move Views in the sidebar, or work with relation Views on Entity pages.

On top of that, Fibery now does up to a third less work behind the scenes, wherever you click.

Spot something sluggish anyway? Our ears are open.

Improvements

• Table View:
  • A click on a rich text cell now just selects it, like any other cell. Click it again or press Enter to open the document, Cmd/Ctrl+click to open it in a new tab.
  • Cells and rows are now selected separately. Click cells to work with values, tick checkboxes to work with records. Shift+click a checkbox to select a span of rows, Shift+↑/↓ to grow or shrink it.
• UI:
  • The Relation settings panel is now resizable in DB Settings → Relations, making it easier to view longer field labels without them being cut off.
  • We recently updated the Automations icon in your Inbox, and now it’s time for a new Formula avatar. Both icons now follow our refreshed visual style and look great together.
• Authentication:
  • Show workspace selector on OAuth screen. When you connect Fibery to another app or AI assistant, you no longer have to type your workspace name by hand. The flow now works just like the regular Fibery login: log in, pick a workspace from the list, and approve access. Typos no more!

Fixed Bugs

• Table View: (Pasha rage-fixed a bunch of table bugs in 2 days)
  • Сlick on Collapse /expand button should be ignored so user can do these actions without navigation to entity
  • Wrong numbers when user adds new entities via New button in toolbar or context menu
  • Counters aren’t updated after deleting entities
  • In table view CMD+Click doesn’t open an entity in a new browser tab
  • Pressing delete in cell where rich-text field is in focus should not lead to attempt to clear it
  • Can not copy cell of rich text field snippet in table view
  • Undo restores only one cell after deleting a cell range in table view
  • To-many relation column overflows into the next row when row height is medium or tall
  • Bad layout on resizing text column when table has medium, tall or extra tall row heigh
  • User gets autoscrolled to the right in Data table on renaming database
  • Aggregated name title in Grid / List should follow DBs hierarchy
  • Error on deleting data from cells if selection contains aggregation cell
  • Unexpected scrolling in the Table on changing ‘Display as’ option
  • ‘Open entity’ button is hidden in the Table row for Middle/Tall and Extra Tall row hights and long entity names
  • Shift+click in table view moves the focused cell to the clicked cell instead of keeping it
  • Wrong expand/collapse state on root items after “Expand All” or expanding via “Shift + space”
  • Wrong counters for Table rows after applying a filter
• Sharing: Level 2 views are not shared
• Fibery AI Agent:
  • AI Agent automations show fields as deleted / phantom broken-formula warnings - missing dynamic samples in ai-agent prompts
  • “Search” tool results don’t contain links to entities/views
• Custom Apps:
  • Custom apps: render mentions of documents and views
  • Hide “Edit app” when user is not in Architect mode
  • Custom app flashes light before its theme applies in Safari
• Documents / Rich Text: Inline comment reopens when clicking text under a resolved comment
• Entity View: Nested menu with buttons: show only when “2 more”
• History & Audit Log:
  • Removed check-boxes are not highlighted in red
  • Owner/Editor is not able to restore space-level document or whiteboard if he is not the one who deleted it
  • Rule name and avatar is missing for changes made by scripted Rules in history
  • There is no records in history if the field was deleted from Enum database via Undo
• MCP Server:
  • Agent sees [Deleted Field] if automation refers to previous step’s collection field
  • Fix most common errors in Fibery AI and MCP
• Service Users:
  • Don’t show navigation button for ServiceUser unit
  • Service User should use the new automation avatar
  • Automation User doesn’t display on views if grouping by Created by field is added
  • No tooltip with User name appears when hover on Automation User in pinned fields
• UI: Double tooltip is displayed for User pinned field

P.S. We hope you enjoy all the releases!

[RonMakesSystems]

So much stuff!! Fantastic. Looking forward to test and try all of this out.

@mdubakov

Build the integration you’ve always wanted and tell us how it went in the community. Good luck!

I personally need to wait until there is an MCP connector to test this out as I am out of Fibery AI credits :///

“Data flows into Fibery only” custom actions can flow data out of Fibery! Custom integrations are great. I look forward to play with and share once the MCP is out.

@mdubakov

Faster Views

@mdubakov

• Agent sees [Deleted Field] if automation refers to previous step’s collection field

Thank you!!

[antoniokov]

@RonMakesSystems

as I am out of Fibery AI credits :///

You probably know this already , but a reminder just in case: you can enable extra usage and go beyond what’s included in your subscription.

[grex22]

Wow… Faster Views is VERY noticeable this morning, especially on Dashboards (with, obviously, multiple views included in one). EXCELLENT work, and I’m sure some very satisfying optimizations went into that effort!

[helloitse]

@mdubakov

You can now narrow search down without reaching for the mouse.

This is amazing! Serious quality of life game changer, like multiple times a day, every day, every week.

[bikesupplies]

The new custom integrations via AI is *!!! Shopify orders and customers synced and new orders arrive within 20 seconds via webhooks

*I’ve never used that emoji in my life. I assume I’m cool now.

[myg_ge]

Is this AI integration also a one-way sync?

[tah4]

Revert the Text Table action menu changes and implement the copy/paste support some other way, please! It was so convenient being able to right click anywhere anywhere to access the Actions menu, now I have to hunt down this minuscule little orange button. Even worse, I can no longer do actions on multiple rows/columns at a time… I used to be able to select a set of rows that have merged cells, colors, etc > clear content, and now I have to individually duplicate each row, individually clear contents, then drag to arrange them in the right order. Wouldn’t it just be easier to add Copy to the action menu?

Please find some other solution and bring back the Text Table features we had last update!

[mdubakov]

@myg_ge

Is this AI integration also a one-way sync?

Yes, but it also can create Custom Actions for Fibery Automations, so with some rules you can have at least partial 2-way sync

[alex]

@tah4

Revert the Text Table action menu changes and implement the copy/paste support some other way, please! It was so convenient being able to right click anywhere anywhere to access the Actions menu, now I have to hunt down this minuscule little orange button. Even worse, I can no longer do actions on multiple rows/columns at a time… I used to be able to select a set of rows that have merged cells, colors, etc > clear content, and now I have to individually duplicate each row, individually clear contents, then drag to arrange them in the right order. Wouldn’t it just be easier to add Copy to the action menu?

Unfortunately, the browser’s context menu includes not only the copy action, but also a bunch of other actions and extensions for different uses. We’ve received several complaints about replacing it with our custom menu. I’ll look into what we can do to improve the multi-select behavior for rows and columns.

[Sev]

Well done again! Thank you.

@mdubakov

Type filters right into search (experimental)

As a team that press CTRL/CMD+K sometime fifteen times in one minute, you cannot imagine how much this will help. With a deeply interconnected dynamic label system, it is really hard for us to find anything without filtering by Database/Type. This is a VERY clear UX improvement for us that will have us shout blessings at you and the team many times a minute!

Maybe if you get get a chance, add shortcuts like #: for public ID, and shorter forms like ppl: for people, and sp: for space, and perhaps on: for created, though that last one is debatable.

You have directly solved [UX/Search] Global Search - Inline Textual Filters Within Search String and Indicate the target database in search query.

@mdubakov

MCP Server can execute code now

This was amazing when I noticed it yesterday when the MCP calls asked me to allow them to execute code. So much more efficient! Can the code blocks do pretty much anything we can do in a Fibery Script Action Step?

@mdubakov

Rich text changes history in your Inbox and entity history

Yesss, also very helpful, presuming that MCP can access it, too, since until now it could not see the actual content changes made to a Doc.

[maxdubakov]

@Sev

Can the code blocks do pretty much anything we can do in a Fibery Script Action Step?

Hi, glad it’s helping you already Not exactly, it is a separate implementation - it doesn’t have args variable, or fibery & utils packages as Fibery Script Action Step does, and is limited only to tools available through our MCP. So, it can do anything Fibery MCP can do

[Sev]

It is great that it is a separate implementation. I thought as much s I looked through the code snippets that Claude was sending to it.

So it is a batch mechanism mainly for the MCP, which honestly makes it SO MUCH MORE efficient. Thanks again.

[Sev]

@mdubakov

in: also works in # and [[ mentions inside documents.

This deserves its own reply, which is what I am doing here.

A few minutes ago I realised that it worked for mentions, too, and it made the feature immediately at least doubly useful. Bravísimo. WOW!

[ruslan]

Thanks for adding rich text history!

Is there an SLA that you aim for with Rich Text History? I use history changes API to perform data synchronization. Sometimes it takes ~15 minutes for the changes to appear. Are such delays expected, or is this something that you plan to improve?

[mdubakov]

Usually history is updated after 10 mins of inactivity in a rich text

[nathan_jay]

Great updates on Rich Text history and the Type filters are handy.

MCP can execute code, I’m not sure when but it sounds like it will be super helpful when it does come into play. Is this heading towards CLI/SDK as the best agent tool?

[nathan_jay]

Also on Custom Integrations it’s a fantastic upgrade.

Now that you’ve made the massive leap to bring Custom Apps and Custom Entity Views it has raised the ceiling for UI. But they’re only as good as the data.

So for established workspaces I think these things will be very valuable:

1. Migration path for the existing Custom Integrations

For example, I have a custom integration and I need near “real-time” in Fibery which is current blocker to moving more work into Fibery. There are no webhooks available (for now and maybe some time) so reducing sync to 5 mins gets me close enough. I can’t rebuild it on this new platform unless the new integration will serve the existing dbs.

It would be bad if there were now 2 worlds of integrations old v new and each with very different capability.

Is this possible now? Or maybe its not even a feature but something you can do in the backend for us?

2. Customising existing native integrations

I have a few native integrations which now have many views, reports, etc. I want to tweak the integration to add a few extra fields to the sync. It’s very unlikely my requests would ever get done to the native integration because I’m only 1 request and it will never get enough votes for the effort.

If I can “take over” the native integration so it now becomes custom then I can build on it myself.

It also opens the door for Fibery to roll out a whole bunch of new native integrations so that new customers have a lot out of the box. But if someone doesn’t like your specific implementation, no problem, they just customise it later.

This solves the 'blank page" barrier for new Architects onboarding. Get them up and running with data actually flowing into Fibery as fast as possible with no friction with no fear of lock-in.

2. MCP support

For all the exact same reasons as Custom Apps and the building/tweaking of integrations mentioned above e.g. using subscriptions. One thing I also found critical to debug data errors and optimise syncing is for the agent to have access to the logs so it can answer things like “Why does this data not appear in Fibery even though its in X source system?” or “My full sync is taking 2 hours can you optimise it?”

@mdubakov

• After that, we’ll most likely charge for compute time: the more your connectors run, the more you pay

Fair enough and I much prefer to have the price/value trade off to decide. For what its worth this is my MONTHLY usage on my largest custom integration by far. Its syncing hourly (delta) and includes a few forced full syncs after schema changes.

[ruslan]

Thanks! Makes sense.

Is there a way to learn which documents are being actively edited now? I have an integration that would benefit a lot from having all the latest versions of the documents, even if they are being edited right now. I understand that you don’t want to pollute the history with constant “document changed” updates, hence 10 minutes debounce window.

I am more than happy to use undocumented API, if it will give me access to this list of documents with pending edits.