(Better late than never in this case)
September was a productive month, but we yet again blurred our focus on many areas. Initiatives assessment:
Granular Access. Had a major release, but the other thing is on hold.- Growth. Not a lot going on lately (but it’s honest work).
Fibery Apps. Good progress and many things were completed.- Make AI Agents happy. Velocity is not high enough
- Other. Looks like too many things here again
It is far from OK, in general we should have less effort in Other, otherwise we are not focusing enough. Overall, I think Fibery as a product is already quite strong and one of our main problem is distribution/promotion/education/etc. We should focus on things more that will improve probability of Fibery to be noticed, especially via ChatGPT or Claude:
- Re-focus on “totally awesome, fast, best in class MCP Server”
- Promote MCP usage, showcase, make MCP use cases stand out
- Templates/recipes, apps gallery, custom connectors gallery, etc. to showcase what Fibery can do in domains we know nothing about (manufacturing? logistic? green energy? etc)
- Plugins with skills to increase chances to be mentioned by ChatGPT and Claude
- Open source Fibery connectors (with good docs, etc)
- Maybe more public things like
Share Apps / Build Custom Portals with AI that can expose Fibery to 10x more people
Granular Access
Hiding data from Admins
The first release of Allow to hide Space data from Admins is out there in the wild!
“Immediately starts POC for Performance Management space
”
As we assumed, plenty of places had to be adjusted — with automations proving to be the most challenging bit. The good news is that we’ve (hopefully) paved the way for Allow non-Admins to use scripts in Rules.
A few smaller things to do before we move on to the next thing:
- Ensure Admins have access to all User Groups — otherwise, it makes things awkward in edge cases.
- Share Databases and Entities with Admins now that it makes sense.
- Promote the option to hide data from Admins because when the tree falls… (you know the rest)
- plus, a bunch of smaller bugs to fix;
- plus, anything that pops up based on feedback.
The first major core Feature in permissions led by Dima. Cool!
Next
The next big thing is Add create capability to access templates with the aim of Dima being even more autonomous from Andrew, including coming up with the architecture of the whole thing.
Growth
AI pricing
Cap AI limit for largest workspaces will curb the enthusiasm of our largest customers who are also the power users of Fibery AI — starting in October. We expect the monthly costs drop because of the caps (assuming other consumption patterns stay stable).
Smaller things
Vova has done a couple of small things:
- Bump monthly usage limits on the first purchase benefits our customers. It was a bummer before when someone purchased a 1-seat Pro subscription and still got stuck because they spent their monthly limits during trial.
- Consume AI credits in rich-text assistant benefits us.
Custom Apps
Done
Quite many things were done to improve Custom Apps platform and development experience, here is the list.
- Save prompt draft
- Support pasting images from clipboard to App agent
- Open Entity View in panel from custom app
- Better rich text component for Custom Apps
- Invent a flow for “platform” update for Custom Apps
- Allow to change url for the App
- Switch on some lib for querying, like tanstack/query
- Encourage to build custom apps via MCP
We’ve also released an ability to to create entity views as custom apps. It enables many new use cases and our users are quite happy with this release → Custom Apps as Entity Views:
Let’s check some metrics. Custom Apps adoption is fast and now they are on track to catch up with Forms and List Views.
More workspaces, more users, more iterations and higher cost for us so far. Architects iterate heavily via MCP, but more workspaces build custom apps within Fibery. At some point we should find out how to earn and not lose money on custom apps. We will observe how things are doing in October and make some decision in November.
Next
First, we will do some minor improvements in Custom App platform:
- Create Apps from Fibery AI
- Add routes so exact page in app has own url
- We will try to make UI in Custom Apps closer to Fibery UI
Second, we will introduce a way to create Custom Connectors via AI fast
Make AI Agents happy
Relatively slow progress here. The most significant things were a complete re-write of Fibery AI Agent and suggestions.
Done
- Access, Create & Update Whiteboard View with AI
- Migrate Fibery AI agent AGAIN (to JITT)
- Add fetchByUrl to fetch public urls in md
- Now agents can suggest changes in rich text fields and documents Suggestion mode for rich text
Next
MCP becoming more and more important for product success, so we should always work on its performance, quality and breadth. Serge agreed to take this area as PO, so he will work closely with Max, Victor and Alexey to make the fastest, cheapest and coolest MCP Server in the next several months.
MCP vs Fibery AI adoption is quite telling.
AI tools for Fibery AI and MCP will be expanded and improved:
- Currently AI makes many mistakes when creating automation rules or buttons, we will fix it with a strict validation and better tooling in Make Automations rule creation via AI more reliable
- Upload local files to Fibery via MCP
- Check field dependency before deletion
- We hope to add proper skills to Claude and ChatGPT that will help architect work with Fibery via MCP, like
fibery-architect,fibery-health, etc.
For the future. We have several major areas here that we did not tackle yet, but have to soon:
- Agents in Fibery Automations and a simplest feature here is Fibery AI Action in Automations
- AI Agents as Users - delegate job to agents and manage it via Fibery (Fibery AI can be an agent as well)
- Workspace Monitoring - Collect and expose activity and health data about every object in a workspace (spaces, databases, fields, views, documents, entities, automations, integrations and users) so that both admins/architects and Fibery AI can see what is broken, what is unused, what is slow and what nobody reads, and act on it.
Other
~30 features were done in various areas, including Rich Text, Whiteboards, History, Automations and Flexible Domain, but I guess we will not bother you with the long list of done features here anymore.
P.S. As always, thank you for your patience and support. It is a pleasure to have you in our community 
This is incredibly exciting!