@snehass This is a really interesting question. Being a QA is hard with all of these mixed tasks, and each requires a different skill set as well. You can still use the generic agent mode in StudioAssist or use any AI Agents you have to do all of this, but it will save you time if you create custom agents with a specific instruction for each task you shared. This week in KS 11.0.0, weâve just launched a new feature called Agent Profile to help you with that.
I believe with your expertise, you have your own rules in mind already for what to automate vs what to keep manual. AI can still find best practices in its knowledge base and help you decide, but with more instructions from your side, which represent your experience and understanding of the project/app youâre working on, it can provide better result.
I clearly understand all of this ( Cost perspective , and hence requested straightforward answer) but if now other tools having AI and agents embedded in them and without any licenses costs , why would anyone opt for Katalon.
Coding is required in both , then why wont people move to other options. Your answer would have been perfect , if there were no agents and AI related tools in tools like playwright.
Katalon average license cost ( real projects would not take the free version ) is 2500 USD per year plus more costs on KRE. and in real projects all QAs would be equipped with the licenses, So cost is involved here as well.
Not sure if its only me who is thinking in the wrong direction. But as a katalon creator , I have been asked this question multiple times and I have no concrete answer
Fallback chain: It helps us predict brittleness before CI. If the âbackupâ locators are things like deep XPath / dynamic classes / index-based selectors, weâll fix the object immediately instead of waiting for flaky CI failures. It also speeds up RCA because we can see what was tried and why it still failed.
Ranking: The most useful signals to see (even for top 3 candidates) are:
match count (unique vs multiple matches)
strategy type (Attributes/CSS/XPath) + simple risk tags (dynamic id/class, index-based, text-based, deep path)
short âwhy this wonâ note (âunique + stable attributesâ, âŚ)
Even a small panel showing Top candidates + risk tags + fallback order would be super valuable.
@duy.lam
When will auto remediate feature come to save the time getting lost in self healing. I am not sure whether the smart locator works like that or not, I tried but I didnât get proper result.
Summuary of the questions: Self healing is good but worse when we execute the regression pack, so once the system detects the useful locator then it should off its own change the weightage of the locator, so that in the nxt time onwards instead of wasting time on self healling it should pick the right locator and either skip the broken locator or delete the locator from the Object repository
@shin
How is the new Test Recording Agent improving locator stability and reducing flaky recordings compared to the traditional recorder?
Most of the time when I use recorder to design the test, I feel the locator getting captured more often are flaky them I need to manually find the dedicated xpath and input to remove the flakiness.
@Shin Katalon studio has serious issue with stale element error, In playwright this doesnât happen.
When we create test and provide some condition to check for 15-20 secs, the why do we get error of unable to wait for 15-20 sec. Then to deal with this situation unwantedly I need to make lots of extra test steps like page load, scroll to element and more. Refer the image
@xuan.tran Will the Test Recording Agent be able to automatically detect dynamic elements and apply self-healing strategies at recording time instead of only during execution?
If this is provided then a big relief for all automation guys.
@shin when will the katalon studio come as web version or even in desktop version, there should be feature of auto update of version. Why still we need to manaully download the latest version and make many installations like to put the jar files like JT400 for db testing for each newer version update.
Like for example if I am using currently 10.2.3 version and everything is working find I see newer version have come in the market, so now to shift I need to do lots of efforts, and the bigger problem is that we need to ask entire team to update the version otherwise, we get error msg that project not behave nicely if we jump from one version to another
@xuan.tran How does the Recording Agent handle complex enterprise workflows like multi-tab navigation, iframe-heavy apps, and shadow DOM elements?
I still struggle with the recorder when the Shadow DOM or iframe element comes
@Shin
Are there plans for the Recording Agent to automatically recommend meaningful assertions instead of only capturing click-and-type actions?
I mean the can the recording agent be smart enough to recommend assertions automatically, rather than only recording raw user actions?
@shin
Can the Recording Agent intelligently break long recordings into reusable test components or modular keywords automatically?
One incident happen recently with me, I had one test scenario where the flow is very complex and lengthy, I wrote approx 700 + test step, and executed. On katalon studio in my local it execute but when I schedule on VM server then it starts breaking and giving me error as method is too long and cannot execute, then I broke that into multiple keywords and then use.
But as AI is getting advanced day by day, the our inbuilt AI katalon studio must have made me cautious or warned me initially.
Studio didn;t give any information about the lenghty method, its when the error start coming, it started saying.
@Shin
Will AI recording support data-driven testing by detecting input patterns and suggesting parameterization during recording? Even the recorder should have some feature where if we want to have locator with placeholder or parameterized then it should record based on the placeholder we give.
In this case we need to manaully create the locator with the placeholder
@xuan.tran How do you envision the future balance between recorded tests and manually engineered frameworks in Katalon Studio? I mean when will the generative AI feature where we can generate the test script based on the manual test be given added with the the katalon studio recoder and we need not to buy anotther TrueTest. When we already purchase the katalon license then everything should come inthat license itself. Right now, KS license, Testops, test cloud, true test,.. list will go on and on.
@arvind.choudhary Thanks for all your interesting questions above. For questions related to the Web Recording Agent, @shin will help to answer.
Web Recorder, TrueTest, and StudioAssist Agent mode are tools we built to help you generate tests, so I believe itâs up to each team and situation to find the balance when using these tools. As software development with AI is moving fast, we also want to equip QE with AI features to help them test at the speed of AI, too. Sometimes, you just need a few clicks using the Web Recorder to quickly generate the test with the exact flow you want, while there might be big or repeated tasks that you will just get requirements and want to translate them into test cases quickly using AI.
For this, I think itâs currently an advantage for individuals and small teams, as they can purchase only the license they need. However, this is off the topic of AKA, so I would not leave further comments about license and pricing here.
âDo we have any plan where Katalon support orchestrating multiple execution agents so that they can cooperate in running different layers of validationâsuch as Web UI interactions, API calls, and Database checksâwithin one intelligent end-to-end workflow where each agent can share data, synchronize steps, and run as a unified automated process?â
@duy.lam
When will search history be available in Studio Assist? Having a record of our past queries would help us refer back to previous conversations and continue interacting with the AI in context, without needing to start a new session every time.
It makes me agitated if I need to research or debug deep down further in any topic which I left open couple of days earlier. Now I need to again begin from start.
@xuan.tran
Why isnât Studio Assist able to automatically look into my projectâs test artifacts and give contextâaware explanations? Since it is integrated with Katalon Studio, shouldnât it detect variable declarations, test suite bindings, and global variable usage without me explaining them? For example, if I ask about a failure in Test Case X, shouldnât it automatically analyze the test case, suite, and bindings to identify missing links or configuration issues?
If the above feature comes in the studio assist then it will be a wonderful experience
@xuan.tran
Will agentic workflows maintain an execution memory across multiple steps â allowing the agent to recall earlier failures, object patterns, or app state transitions to inform future decisions?
As MCP servers extend StudioAssistâs capabilities, will future releases allow custom MCP server hooks for advanced tasks like live DOM introspection, codebase search, or external requirement mapping (e.g., Jira integration)?
My another question is around Autonomous Data Binding Validation
Will StudioAssistâs Agent Mode eventually be able to verify data binding correctness across test suites â such as checking variable type mismatches, missing dataset bindings, or unused parameters â before execution?
