General question: Am I the only one feeling the general development of KNIME AP is stopped? The only progress you can see since several releases is AI-integration and the “modern UI” (which is terrible on Windows btw). Last one is needed to provide a fully cloudbased solution with integrated editor, in other terms to monetize the software (already announced here). But I do not see any basic nodes getting an update, e.g. bringing some similair nodes to the same codebase. Or any request in the “Feedback & Ideas” Forum is getting implemented. I’m just curious if KNIME is working on the “perfect cloud solution” while loosing touch (and with it also users and customers in long term) with its base.
4 Likes
there was some updates to the Expression node (still missing 1/3 of the functionality the Column Expression offered), and some Datetime nodes.
some part of it is limited to the Java version in the background. with the recent update, the jump was made to the 2 year old LTS version 21 (current LTS version is 25), but that matters mostly for “under the hood” performance and features.
as you already mentioned, the big push is going towards AI features, which is a gamble as might become deprecated again in the next 18 months. question is, if by then any “data engineers” are left either.
Most long-term users picked up Knime for the feature of “self-documenting” and automation.
The self-documenting part is becoming less and void given more and more “standard operations” are not made available as nodes but instead its “hidden” inside coding nodes (Expression, Column Expression). That has the “added” benefit that you can have an LLM write code for you in theory, but anyone that thinks about this for 2 mins should become aware that for a coding approach, they are better of going plain python, r, julia or even matlab (or octave for the FOSS version).
Some people want to “build” their data pipelines but are limited (or frustrated) by incomplete integrations (hello path datatype, hello SharePoint lists, hello Parquet files). AI also isnt building and connecting the blocks yet.
At the same time, creating python scripts (hello LLMs) and running those (e.g. with UV) has become so easy and universal that this also seems to be a reason for neglect.
lastly, the communication of the most recent 2-3 years was essentially: we focus on porting everything to the new webui to clean up legacy stuff and become quicker in developing new things.
1 Like
As an enthusiastic KNIME user, I see it similarly: the lower the barrier to hard coding becomes thanks to LLMs (or rather, no longer exists), the more the need for low-code tools such as KNIME decreases. A KNIME MCP interface could be a game changer, i.e., building KNIME workflows directly in a KNIME web UI with Claude. (n8n already has this Claude connector.)
1 Like
Thanks. Exactly, the “Expression Node” is one of the few examples with an huge step. But as you and Christian_Essen mentioned, it’s not necessarily a step forward in terms of Low/No-Code.
If I want to use LLMs or AI Agents to do the job, there are better tools than KNIME. So this direction doesn’t make sense. Also adding MCP is maybe not “the” way to go (I’m not deep into that). K-AI was a try to step into this direction. Maybe a LLM writing the XML of “workflow.knime” (which sits within each KNWF File) and then the configuration XML of each node in the workflow.knime would also be a possibility. For this the XML documentation must be “known” to the LLM.
@DanielBog mentioned somewhere in the forum, that they are currently working on a different area of the product, but as soon as this is finished they will focus back on improving existing functionality. Maybe he can comment on this.
1 Like
afaik this was mostly referring to the webui port
1 Like
That’s correct - Quality of Life developments (reacting to feedback and ideas, progressing expression node etc.) apparently are planned for early this year.
If I look at the change log of 5.10 a lot of Modern UI porting has happened so hopefully this is either complete or close to complete so that some of the “other” topics can be tackled next!
Overall I think development of the product has been focussed - I assume for a good reason.
What I think we cannot forget is that once you are PE/VC funded, pressure rises in terms of meeting business plans etc. I take that the big enabler to meet goals is Modern UI / Browser-based workflow building (as demo-ed at DataHop Stuttgart) and offering KBH as SaaS - so although I also might prefer prioritising certain features / issues that I see, I can understand why in the broader context the priorities are as they are…
1 Like
This sounds like K-AI “Build” feature?
Hi folks -
Well one thing I can tell you is that development on KNIME AP has certainly not stopped. I was just sitting in on an internal call this week where the devs were demo-ing features that are coming in the next two releases 
I will say that we are in a bit of a transition period, both for KNIME as a company and for KNIME Software. As some of you noted, quite a lot of our development lately has been in support of moving to a cloud based solution. Let me see if I can get someone with a better perspective than I have to comment more on that.
As part of the transition, I can say that we are working on a community relaunch that is coming… soon-ish. I don’t want to give away too much on that front yet, but we have some exciting stuff lined up for you there.
Thanks for your patience! We know you love KNIME as much as we do. One of the core principles we talk about is “Don’t Break KNIME”, and that’s at the front of our minds right now as we go through some change.
8 Likes
Nearly, but not the same. K-AI is inside KNIME and step-by-step while a LLM-XML would be a “workflow at once” from outside.
but we have some exciting stuff lined up for you there.
maybe its just me, but I don’t care about the “new” and “exciting” stuff, if the basic features are neglected, dysfunctional or inappropriately complex.
3 Likes
You are not alone! Thats the reasin I made this posting.
2 Likes
Well I guess you are right in that the LLM puts one node on the canvas after the other - but surely not only one per interaction (i.e. K-AI can build workflows of multiple nodes).
My guess is that in the backend this is actually being done by creating the XML structure - probably with at least some KNIME-specific RAG to assist the model with making less mistakes.
I have not tried to see if e.g. Claude Opus / Sonnet 4.6 are capable of providing a full .knwf that actually works, but seems like people have tried:
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.