Dinakar
All right. Hello everyone and welcome to Pcloudy's webinar on how QA teams are saving hours every sprint using Claude plus Qpilot MCP. My name is Dinakar and I'll be the host for today. before we get started, I would like to go over a few housekeeping rules. You may have noticed that your line is currently muted and if you have any questions or clarifications, you can use the Q&A option given at the bottom of the screen. We will answer your questions as part of the Q&A section towards the end of the webinar. Also, please do note that we are recording this webinar and we will share the link of the recording to watch again or share with your colleagues later. With that said, let me
Dinakar
Introduce the speaker for today, Jessica Selvin. Jessica is a lead solution engineering representative at Pcloudy. She has around five plus years of experience in the IT industry and prior to her role at Pcloudy she has worked with LTIMindtree as a pre-sales consultant in automation solution design for various industries. She has a keen expertise in data analytics, machine learning and various artificial intelligence tools. So without further ado, let me hand it to our speaker.
Jessica Selvin
Jessica S. over to you. Yeah, thank you so much Dinakar for that introduction and I hope my screen is visible. Yeah. So talking about I mean as Dinakar already introduced our topic for today. It's about how Claude and our Qpilot MCP connection would help you generate tests faster, have a single plane of view for all your test results and basically help you save up those hours with switching between right. So firstly let's just look about look at how AI now what are the problems that AI has solved
Jessica Selvin
And why are so some things still taking more time right so for example test generation if you notice with the way AI has come up is mostly solved however whether you're using an AI agent like Qpilot or you're using your GitHub Copilot and writing your Playwright scripts all right you can describe your flow and you will be able to get your scripts pretty quickly. right and that's genuinely impressive and it's been commoditized pretty quickly and the market has it right and however most QA teams are still using several
Jessica Selvin
Tools right so for example a QA engineer could touch between writing a test and then sharing the result and then they're they're using an AI assistant and then they have their editor or IDEs and then their device farms and then maybe dashboards and reports in a separate place and then logging bugs or tasks to track on their Slack or Jira. right so there's a lot of context switching going on there. and additionally if there's API testing then maybe a Postman as well to do the API tests and there are at least three or four manual
Jessica Selvin
Handoffs that take place to debug a single failure. All right and all of those need to be logged in separately in those individual tools. right so that's where you would see lot of time hours lost in every sprint or every release cycle right where you would be context switching between these tools and often times that particular details are lost between tools so the full context is not there when you're feeding it to your AI tool right and again manually looking at logs manually looking at dashboards and
Jessica Selvin
Reports Right? Whereas most of your work and innovative work is happening in your AI agent like your Cursor or your Claude Code right you're not able to get that context back into the tool that you're primarily using to develop your scripts generate new builds and etc. Right? So that is the big the problem statement that Qpilot and Claude MCP connection seeks to resolve.
Jessica Selvin
So just let me summarize the u key agenda for this call. right firstly you'll know how Claude and pupil MCP connects. All right and you'll directly be able to access our real devices from your tool be it your Cursor or your Claude. All right. And secondly we will have a live demo and walk through of this. thirdly, I will also cover about our AI power test failure analysis reports and how debugging with our AI failure analysis is highly cut short in terms of time where you'll be able to look through the
Jessica Selvin
Logs and get detailed analysis quickly. right and how you'll be able to execute your test cases, how you can have a reusable MCP setup for your team and as mentioned yeah we will take up questions towards the end of the demo. So let's look at section one where are QA teams still losing time why are they still losing context between tools.
Jessica Selvin
All right. So if you look at today's workflow often times you would start with your AI assistant. It could be Cursor GitHub Copilot or it could be your Claude Code right where you're you describe your flow. You're trying to get a test and that back and forth happens and then finally you have a ready to run script. Right.
Jessica Selvin
Secondly the editor you copy paste that script out of that chat window and you paste it into a code base. All right. And you fix those imports and maybe there's some page object conventions and locator swaps that you need to do. All right. So here there could be some back and forth going on your IDEs or test editor and finally on your device farm. Right now you will be able to book a device then you will take you will add your capabilities connect to a device and then go ahead and run your test cases. Right? So you'll be able to track it. you'll be
Jessica Selvin
Switching between your IDEs or your CI/CD tool and the device farm like Pcloudy right and then your dashboarding for your results right so you might have developed a separate standalone solution for your dashboarding right and you might open that up and you will be observing your results and finally any documentation or any updates back to Slack or Jira would be done where you're updating bugs Right. So here you can see we have switched between five tools and you have five chances to lose context. Right? So
Jessica Selvin
You would be losing critical details or bug fixes that were done on your test editor. right you may not be able to get that context back into your AI assistant to further improve your accuracy in terms of your skills. Right? So that is the context that you're losing which eventually ends up in losing a lot of time right.
Jessica Selvin
Yeah. So again the other aspect so we talked about switching context the other aspects where time would could be getting taken up is in the booking of devices uploading of bills tracking your execution status. All right. So every time you have to go to the dashboard or go to the active session, see if the live session is done, see if my report has been generated, right? and then it might look like a simple thing, but these kind of tasks do end up eventually piling up and taking up time and next is digging through raw logs.
Jessica Selvin
All right. So, any test case that generates your logs and if you want to see what went wrong in your test case, you might have to go through those logs and that takes up time. With our AI failure analysis agent, you will be able to see the relevant log lines where exactly there were failures in your test case. And finally, the report, right? So if you have a separate reporting tool for your test cases again pulling that up and seeing if those results are actually reflecting all of that might have some inbuilt latency to it right and also being able to fetch it and
Jessica Selvin
Debug and then report it back into your Jira or Slack right so all these touch points are where you would see time and context leaving. So next introducing our Claude and Qpilot MCP. it's mentioned Claude here but because that's the focus of our demo.
Jessica Selvin
However, we do have integration with other tools as well like Cursor and so on. So what is MCP? right it's Model Context Protocol if you remember it made a big buzz in market few months ago. right and it's an open standard right from anthropic and the honest oneline version is this right it's a way basically for your AI assistant to talk directly with an external system so instead of you being the relay between them right so it directly is able to fetch data from your tools and also perform action on your platform right so that's the whole
Jessica Selvin
Idea right that's the core idea of an MCP connection and it becomes common protocol to interact with different tools. All right, so that it simplifies this connection. So that was the key selling point of the MCP solution. All right, so basically it's something like an API connection between Claude and your device ploudy where you read from one retype to the other and carry the context in your head. Right? So MCP takes you out of the loop and directly talks to your platform. And coming to Qpilot, if you've not already heard
Jessica Selvin
About Pcloudy's Qpilot, it is our AI based test creation tool. you can check out our other webinars on Qpilot or look through our website to see further demo videos of Qpilot or you can reach out to us. So basically it's an AI powered test creation platform that sits on top of Pcloudy's real devices right so you might be aware that Pcloudy is a device farm and we host real Android iOS devices as well as a browser Claude for you to run your mobile and web test cases be it manually or be it through your Appium Selenium Playwright or
Jessica Selvin
Other popular frameworks that you might have or through a our Q- pilot layer which is our AI layer where you will be able to give plain English language descript descriptions and it will convert them into a set of automated test steps. Right? So that's our Qpilot solution. And then finally, as I already highlighted, our real device infrastructure is the core of Pcloudy solution where we are hosting real physical devices. So these are not simulators or emulators on our device Claude. So you'll be able to access all these capabilities, our AI layer as well as our core infrastructure layer of real
Jessica Selvin
Devices. with claw. So what can you achieve? Right? I'm just trying to point out some of the most common or feasible test cases that are possible that you can achieve once you have this connection up and running. Right? So firstly generate tests. You can simply instead of going to our Qpilot agent you can go to your Claude and once the Pcloudy MCP connection is enabled you can simply type your descriptions ask it to create a test case ask it to fetch a test case right you'll be able to do all of that you'll be able to retrieve results report and also the live
Jessica Selvin
Execution so if there's any live connection that you've done or a even a manual connection you'll be able to pull up those live views and the reports and because of which you can easily analyze failure and put in that context into your Claude directly. So you don't need to copy paste log lines and you don't need to give feed Claude with further context. It will already be able to pick it up from the reports right and then you'll be able to orchestrate and run collections.
Jessica Selvin
So you can run your u multiple set of test cases directly from the Pcloudy. So next coming to the workflow and the fix right. So how it would be without MCP and with MCP right firstly in terms of the context switching right so you would have five separate tools for one task so it becomes one conversation for generation another one for results another one for analysis and reports right and for booking the device so all of these things would become separate siloed activities instead of that when you have a device booking and execution track by hand
Jessica Selvin
That already solves part of the problem, right? So, execution is still triggered from Qpilot. So, what changes is when you stop doing that and you do the end-to-end solution directly from your MCP tool, right? Next, device booking and yeah, next the failures and debug debugging log by log, right? So, that is again something that's going to be eliminated when you have our Pcloudy AI failure analysis. And finally you will have a shareable report link that gets generated and Claude would give it to you and you'll have all that context of the logs
Jessica Selvin
And failures fed back into Claude. Right? So these are some of the key advantages that you would have when you are running your test cases creating context and triggering your even your manual connection or booking a device anything right and even simple queries like that might take up some of your time right like say I want to find out if a certain device is available you can just say hey list out the Android 15 devices or list out the Google 14 version 14 or version 16 devices Right?
Jessica Selvin
And it'll pull out what are the available devices. It'll pull out what are the applications that you have. You can ask what is the latest build that I've uploaded on Pcloudy platform. So you don't even need to log in each time to fetch basic information about the platform or about the real devices. Right? All this saves up on your time and helps you quickly get started with your testing instead of going to the tool every time to fetch even basic information. So next talking about our AI powered failure analysis reports.
Jessica Selvin
Right. So this takes your raw logs into root causes automatically, right? One at the test case level and even at the build level where you will be able to have full context about your any failures about any element not found or even performance issues if there was any lags, what could be the possible cause of those lags or latency in your test cases. right and as well as recommendations and suggestions for a fix as well. So all of this is powered by AI and you get this report default for every single test case run right so you will get a build-level analysis as
Jessica Selvin
Well as a test case level analysis. So again how this would play into with Claude and MCP right so firstly how does manual debugging work right so you might open your dashboard and you find a fail run and then you go ahead and cross reference the logs against the screenshots by hand or you look at the video and fetch that and then you work out which frame goes on with which log line and you form a theory about the root cause. right and that could be u that could take up a significant amount of time and then you might have to collaborate with different teams you
Jessica Selvin
Don't have a single source of truth right and that could take up a whole day right or a whole afternoon trying to debug that right and on the other hand how would it improve with Claude and Qpilot mcp right so Claude will retrieve the failed-run logs and the screenshots you don't have to fetch it and it would read it and because our p-Claude reporting also has AI inbuilt with the root cause and the log lines shortlisted. It will easily be able to give you further context for you to debug further right and that's the part which matters because the answer really lives in the relationship between them rather than
Jessica Selvin
Them working in silos. So the correlation between the data right which takes up a lot of your tedious manual time right and switching between tools. So that's the place where it would really come and bring in that benefit, right? And you also get a shareable summary. You can fetch that for a later on if a test case fails, you'll be able to link it with a previous error. So you would have a whole history of data and context for you to easily debug as time goes. So with that, let me get into the live demo of our Claude MCP solution. So before I get into the MCP I just for
Jessica Selvin
Folks who yes I hope my screen is visible. Yeah. So with before getting into the Claude and the MCP connection I want to give an introduction to Pcloudy in case if there are anyone here who are unaware about Pcloudy and who we are. So we are a device Claude with intelligence right and this is and we're a SaaS-based platform that can be accessed from anywhere. All right. And the core of our solution as I mentioned is that we host real mobile devices and browsers. We have several data centers across multiple regions where you will be able to connect to a device, launch your apps, launch your URLs and do your testing, right? So we give you the test
Jessica Selvin
Infrastructure. This is the landing page of our platform. Now coming to the real devices and browsers that we have. So all these devices that you see listed out are real physical devices hosted across different data center locations. You can see the different locations here. right and you can upload your builds and get started with your testing. It works similarly for our iOS devices as well. And again we have a browser infrastructure as well where you we have Windows and Mac machines where you can connect to different browsers and different versions so that you can do your cross browser testing as well. And additionally we also have
Jessica Selvin
Support or integration with all the popular automation framework tools. So we have integration with Appium Selenium Playwright and several other tools that are listed out. You can check out our documentation on it as well. And now coming to the Qpilot solution that we have. All right. So this is our Qpilot solution where you can create automation test cases with just plain English language description.
Jessica Selvin
Right. So you can see here. So you can do all this with our UI. But what my focus would be is that how you can do these and create these scripts even without logging into Pcloudy. Right? So but to give you a description about what Pcloudy is right we sorry what Qpilot is that we are it's an AI based test creation agent. So you can see the plain English language descriptions here. right where you can just define it in you don't need to know any keywords or anything as such.
Jessica Selvin
You just give plain English language description and it will create it into a end to end test case. Right. And again we have support for web, iOS, Android as well as APIs. So you can test all of this in a single platform. So Qpilot in and of itself consolidates several tools right. So you might have a separate API tool, right? You might have a separate device farm but with Qpilot because the backing of our Pcloudy real devices is there along with our AI agent and along with reporting and integration with Jira, Trello and Slack and as well as API testing now available as well. So
Jessica Selvin
You really have a lot of tool consolidation with our PCI solution as well and adding to that with MCP you will be able to reap those real benefits. So now coming to the Claude MCP part. So here is my Claude account right and we have a documentation on how you can enable MCP connections. So you can see here if you look at my connectors here listed out you can see there's a Pcloudy toggle that's already enabled. So what this means is that there is now an active connection and Claude is able to talk and fetch details and information from the Pcloudy platform both from the real
Jessica Selvin
Device infrastructure point of view as well as from the Qpilot test create creation part as well right so let me just give a very simple description or an ask maybe I'll say list the available Google 15 version 15 and version 16 devices. Right? So I have just given a plain English language description on Claude and it's right now what it's doing is it's trying to find the tools and get the device list. So all these are requests that are currently going on
Jessica Selvin
To Pcloudy right and it'll get the relevant context and you can see here it's listed out the available devices right so maybe I can just similarly list available iOS devices I can just do that and it will go ahead and pick out the available iOS devices and give me the list Right.
Jessica Selvin
So you can see here it is it will go ahead and list give me the different versions and configurations. Right. So you can for example if you want to know whether a particular device is busy or available or you quickly want to know which devices I can use for automation you can simply list it out over here.
Jessica Selvin
Right. And you can even book a device. Right. So I can say for example connect to a Samsung 14 device and install the IKEA app. Right? So I can just give this description and it'll go ahead and connect to it.
Jessica Selvin
So yeah, some amount of time is expected when it's talking to the Pcloudy tool. right. So however you can see that it's the each API is being called and each actions are happening behind the scenes while the device is getting booked. So you can see there are multiple actions that were performed.
Jessica Selvin
So let me just go back to the Pcloudy and see the active sessions. You can see that it's already picked up the device and the connection is ongoing. So I hope this gives you the an idea about the kind of interactions you can do when you have your MCP connection up and running. I also would like to show you how it [clears throat] work with RQ pilot agent. Yeah. So you can see here I've given it a prompt to create a new test case. So I've given the test case name. I've also given the APK/IPA u name as well as all the details required
Jessica Selvin
To create a test case. And I've also given which device I wanted to generate on the Google Pixel 2 Excel. Right? So you can see here that it's mulling over it and each at each step it's performing the action. Right? So it's found the demo project first and then it's found the folder and it's got the folder ID and within that it's going to go ahead and create the test case. All right. So once it has everything it needs it will actually go ahead and perform the action. And you can see that now it's created a test case with the session ID and it'll summarize what all the actions that it has performed. So
Jessica Selvin
You can see it's also prompting you for the next steps. It's asked can I go ahead and add the more steps to create the test case. So now I can give that. So you can see I've just added plain English language description and you can see all the seven steps are just getting added and getting created in this particular test case.
Jessica Selvin
So you can see here that it summarized the findings as well. It ran end to end without any issues on the Pixel 2 XL. You can also go ahead and edit an existing test case. So for example, if you just want to add a couple of steps on an already existing test case, all you need to do is simply say connect to Excel device and updates these steps on this test case, right? And it will go ahead, open it up, add those two test cases and finish the test case. So you'll be able to easily update your test cases as well.
Jessica Selvin
So you can see now if I come to the Qpilot side you can see that the test case created with the exact name specified on the Claude and the test automation test script has also been created. So now I can simply run it. I can run it from Claude. I can run it from Qpilot.
Jessica Selvin
Right? I can just go ahead and execute these test cases and get the results as well. So you can see now the test case is ongoing and you will be able to see it live and you can get the links to track this live even on Claude and then you will get the reports as well to track it. So coming to the failure analysis part which is the second aspect of what I was talking about with in terms of quick analysis with our AI failure analysis agent right so any test case
Jessica Selvin
Or any automation that you run you will have detailed automation reporting in the Pcloudy platform itself. So you can see here what is the latest run that was executed and I can simply click on the build and you'll be able to see the status of all your test cases how many completed how many failed how many timed out and how many are in progress and you'll also get a build-level analysis right so you can see which sessions had an element not found error or which sessions had a slow page load right and which devices had issues right so you can see From all
Jessica Selvin
These runs that we've done in the past, you can pick out which devices had a higher tendency to fail or if what was the error. All right, so you'll be able to debug all of this at a build or your project levels. And you can also get into any individual test case and debug as well. All right, so you can see here in this case I have the text logs which is the output of each and every step.
Jessica Selvin
You'll be able to see what was the clicks done and then you'll also have something called a step view, right? So you can see each and every step and at which exact step did my test case fail and with what error right and if you want to go deeper you we do give you access to the whole raw logs so you will have access to the entire logs however with our AI analysis that would be fasttracked right so you can view the entire raw logs as well as your device logs here and then we have our AI analysis right so you can see here in this case the test case failed with an element not being found. it
Jessica Selvin
Corresponds to the error detected here in the step view as well and it's giving me the relevant log line as well. So what is the exact log line where it found that particular error and also suggested action. So first validate this particular elements XPath. So you're able to directly pinpoint that okay this clicking on this particular drop-down option for selecting location that particular XPath was probably faulty and that's why my test case failed right so you're able to pinpoint on your root cause quickly and you don't need to spend time looking through
Jessica Selvin
Your logs and as of course we do give you access to the entire video and screenshots and all of that as well and logging bugs Also, you'll be able to log it seamlessly from the platform. So, we have integration with your Jira, Trello, Slack, GitHub as well. And you can simply log your findings directly from here from your Pcloudy platform and these reports are also shareable, right? So, which is why it's much more easier to give context back to Claude. these reports you can just simply click on share link and you'll be able to share it with your colleagues. even if they're not
Jessica Selvin
Registered on Pcloudy they will have full access to these reports. So that is a quick glimpse on our failure analysis aspect as well. So that's about it with our live demo. I hope some of the key aspects that we have that I showed you live has been covered like how I do a real user flow with Claude pulling out results and reports all right you can further ask Claude to analyze your failures as well and you'll be able to see the full workflow in one conversation all right from your
Jessica Selvin
Booking of the device to generating and creating your test cases to the results and analysis. Now, how do you get started with this? Right? So, it's a very simple setup. All right? It would hardly take a few minutes to get started with your Claude MCP integration. Right? So, you would first need to have a PCI account and whichever tool, right? Be it Claude or Cursor. You just need to configure the Qpilot MCP server and then connect that to Claude and then you are good to start generating and analyzing results. So we have a full
Jessica Selvin
Documentation on this along with the prerequisites. So we'll be happy to share that across as well so that you can as in when you like to get started with your MC connection with your Pcloudy account will be able to do so. So you can see here this is the documentation for configuring your MCP server with Pcloudy and the prerequisites are also mentioned along with the exact steps on how to do so.
Jessica Selvin
So I hope that gives you a good understanding about what you'll be able to achieve and what kind of optimizations you can do with having your MCP up and running with our Qpilot as well as Pcloudy. right and feel free to reach out if you have any more questions or if you need a further demo or clarification.
Jessica Selvin
Yeah. So with that we can take up any questions that are there. now on the Q&A side I'm not able to see any questions. so you can add any questions that you might have on the Q&A tab. Yeah. let me just Yes. so I see one question come up. do I need to know how to code to use Claude plus Qpilot MCP or can a manual tester use this true? no a manual tester also can definitely use this first of all Qpilot itself does not require any coding knowledge. So and on top of
Jessica Selvin
That to enable an MCP connection if you follow the steps and once your connection is through you don't need any coding skills for this right it's just a matter of connection and you can go ahead and do both your automation as well as your manual testing without having any coding skills as such. How long does it take to get the setup?
Jessica Selvin
Yeah, so as I mentioned it doesn't take long at all. It's just a matter of few minutes. In less than half an hour, you will be able to get going and create your first test script itself with Claude MCP, right? Does this work with our existing test scripts or do we have to start from scratch? No. If you already have your test scripts on Qpilot, you simply need to go ahead and you can fetch it, you can edit it or execute it directly. You don't need to start it from scratch or only once your MCP is enabled. even test cases created before that will be can be fetched by the MCP. Does this
Jessica Selvin
Replace our test automation framework or sit go along with it? So one is if you're using Qpilot then you'll be able to seamlessly just run our Qpilot test cases with MCP right if you have your own framework we again have same seamless integration with that as well.
Jessica Selvin
So be it Appium or Selenium or Playwright, you will be able to run these test cases on our devices and browsers. So it works both ways. So even if you have your own automation coverage and you're having your own framework that also equally works right. However, if you don't then we have our Qpilot where you will be able to create your automation test. Yeah, I think that's about the questions. I don't think there are any more. Feel free to reach out to us further you can go through the demos of Qpilot or if you want to live demo with us
Jessica Selvin
Also you can reach out we'd be happy to
Dinakar
Guide you on it. All right, agreed. thank you so much, Jessica. Thank you for, sharing those valuable nuggets of information about, the Claude and MCP servers and how we can use all of that together along with our Q- pilot on the Pcloudy platform as well. And yeah, thank you.
Dinakar
We also want to thank each one of you, each one of our participants who took the time to join for the webinar and be a part of it. I'm sure that some of the ideas or information that Jessica was sharing has sparked a few ideas in your minds. my recommendation would be to mull over these ideas. go back to your teams and see how you can implement them. And if you do have any more questions like Jessica mentioned, you can always come back to us, reach out to us through our chat option on the website or write to us at info@Pcloudy.com.
Dinakar
And so do feel free to reach out to us with any questions and we'll kind of respond to those queries as soon as possible. So thank you for once again thank you once again for being an engaging audience and happy testing.
