You guys continue to kill it on fundamental functionality and design of the core app, but still a lot of friction for my team with some basics that would be great to see addressed that just make it harder to “dive in and go” with Fibery:
This one isn’t documented anywhere I think - when creating an entity inline, it still takes 7 - 10 seconds for it appear to be linkable - this is important in meetings when I am taking notes and create the entity inline on the fly - which in and of itself is a feature no other app on the market can match because I can immediately edit fields like “assignee” and “priority” thanks to the amazing foresight you guys had to build customizable displaying of links to entities - but then if I go to a project I have open in my other panel, the entity isn’t there yet…and the meeting participants - we lose flow waiting for it to show up and I have to grab the link…seems small but something that happens 5 - 6 times per meeting
And a big off the topic of this stuff, but It’s been very key for me - I think you guys should consider pushing harder on the AI - I have been using it increasingly and just doesn’t seem to compare in terms of speed and ability to other apps out there now. You no longer show which AI is being used - I recall some months ago (or further in the past) I could choose - so I’m assuming you are picking under the hood now? I am an OG, started in 2019 so there is a ton of stuff I’d like the AI to reference and help guide my business, but it seems to frequently miss stuff - today for example I asked it to find a task about something that was clear in the task but I have a few types of tasks and on purpose I tried to test it to see if I could just find that out with natural language, but it could not…
I do think the theme is an improvement so glad to have it - just wish we could get some “quality of life” stuff that’s been a bother forever knocked out, too! Many more such items on my list but we’ll talk about those later.
Hi @Oshyan, I’ve connected Claude to the Fibery MCP server, and I use it heavily. It works great with everything, very high success rate. Have you tried that yet?
Not yet, though I plan to. I did create a custom API-based integration with my own custom-coded assistant tool before the MCP was available and it showed some promise. But it was a bit heavy and I’ve set it aside for now. The MCP server seems like a good way to experiment more with it.
That said I was more curious if people feel there are off-the-shelf tools (knowledgebases, etc.) that actually do a consistently better job than Fibery’s built-in AI. I’ve heard equally good and bad things about Notion AI, for example. It seems some people have great experiences, but some have quite bad ones. Since I prefer Fibery anyway and most of the AI functionality is paywalled on other tools, I don’t have the chance to compare Fibery vs. other AI capabilities myself most of the time. I’m interested in other people’s experiences around that.
BTW, overall now we are thinking that instead of developing our own agent we will invest more into MCP. I personally believe Anthropic or other vendors are more capable doing agents, while Fibery should be AI-Agents-Friendly, but not work on own agents
I’m already structuring orchestrators and skills to work more efficiently, but I’m still not sure how to bring them into Fibery or how best to interact with them. Right now, the only approaches I can imagine are through Make or n8n.
What I’d love to have is the ability to treat an “AI user” just like a human in a Fibery kanban process. If I assign a task to the “AI user” in the same field where I assign real users, that assignment would trigger the relevant skill and LLM behind the scenes.
With that in mind, how do you see agents being invoked at specific steps in a process, for example? How are you thinking about leveraging the recent AI changes and their impact on how we work and interact with Fibery?
As a workaround, I’ve had to move some of my knowledge into GitHub (even though I’d prefer to keep it in Fibery), because it’s currently easier for my skills to consume it from there. And I heard that via MCP it consume more tokens and also sometimes don’t get the entire context/data.
Yes - MCP consumes more tokens because of how it works (it has to ‘read’ a lot more context). If we could just connect our AI agents in Fibery someway and then point them at the right data, that would be amazing. It would also allow us to control how much we want to spend on AI without breaking Fibery’s bank.
I’m glad to hear. I distincty remember me saying something along these lines, “Enable us to easily connect to an MCP server to get to our data. Enable us to send alerts and change notifications to webhooks. Etc. Anything that makes Fibery a perfect data backend for the vibe coding market that’s exploding right now.“
- we lose flow waiting for it to show up and I have to grab the link…seems small but something that happens 5 - 6 times per meeting
