AI-Powered Test Automation with Playwright MCP, GitHub Copilot & PyCharm IDE (Playwright + Python)

Okay. So in today's class uh we'll see how we can integrate AI so in playright automation. Okay. Playright along with

automation. Okay. Playright along with Python automation. So before jumping

Python automation. So before jumping into the concept so we need to understand some basic concepts like what is artificial intelligence and uh what is generative AI which is basically

subsection of the overall artificial intelligence and we'll try to understand what is generative AI and after that we need to also understand few terminologies like what is LLM which is

large language model the second thing is agent what is an agent how exactly we can use with agent and the third thing is MCP so model context protocol. So

these are the three important uh things we need to understand before jumping into the automation with AI. So first of all uh AI is basically artificial

intelligence. Everybody know that

intelligence. Everybody know that artificial intelligence.

So artificial intelligence is basically as a big umbrella under this there are so many uh subsections are there like we have a machine learning, neural networks. Okay. And then uh we also have

networks. Okay. And then uh we also have a robotics. So these are all comes under

a robotics. So these are all comes under the artificial intelligence only. Okay.

The main thing which we are interested is generative AI. Okay. So this is the main

generative AI. Okay. So this is the main part generative. This is main thing

part generative. This is main thing which we uh important in software development as well as in testing. So

artificial intelligence is a big umbrella under which there are multiple domains are there like we have a robotics and we have something called neural networks data science all these things comes under the artificial

intelligence only but the generative AI is also one of the part in the artificial intelligence where the developers and testers will focus more on the generative AI. So what is

generative AI basically? So generative A is basically generate the content. So

based upon the knowledge which you already have. So for example generative

already have. So for example generative aim means a system where we will train the data. Okay. We can simply say it is

the data. Okay. We can simply say it is a cisive system or it is a framework kind of thing where we can train with some data and based on the data it will

generate the new content or new set of data. So that is basically taken care by

data. So that is basically taken care by generative AI. So the name itself says

generative AI. So the name itself says generative means it will create some new content based on the existing one. But

we need to train it first. So until

unless we train it, it doesn't know what it has to do. Right? So we need to train the system first and then accordingly it will generate the content for us. So

that is basically generative AI is focusing on. So generative AI is

focusing on. So generative AI is basically a system is a system which has a trained data. Okay. Which

has a trained data. So based on this uh trained data it will work basically.

Okay, train the data.

So if you have some data, so we can work with the generative AI tools. There are

any number of tools are there. So we can just use them. Basically how exactly generative AI tools work based on the data which we have trained with them before. Okay. So in this generative AI,

before. Okay. So in this generative AI, this is a very high level terminology is a generic uh term under which there are multiple things are there again. So the

main thing is LLM which is basically large language model okay and the second thing is agent

agent and the next thing is MCP. So this

is very popular model context protocol model context and protocol. So these are the three

and protocol. So these are the three important things we need to understand before jumping into the automation. So

generative is a system which has a trained data. So based on this

trained data. So based on this based on the data it will generate the new content. So the

content can be a text or images whatever it is depends on the data which we already trained based on that it will create a new content. So which is basically focusing on generative AI. Now

there are three important things large language model agent and then MCP. So

let us see what is exactly LLM means. So

LLM is basically a system. This is also a system. As I said, generative AI is

a system. As I said, generative AI is basically system which has a train data.

Right? So LLM is also same thing. LLMs

are examples for generative AI.

Generative AI is a theory concept. But

there are certain tools which are implemented based on generative AI principle. So those tools basically

principle. So those tools basically called as LLMs, large language models.

So we already know there are many LLMs are there in the market. So we have something called chart GPT. You have a Google Gemini cloud deepseek. So these

are all comes under different type of LLMs which are already currently available in the market. So what exactly these LLMs will do? We can do we can ask

the questions and we can get explanation for those uh questions and also we can generate the code also. Sometimes we can generate the scripts automation script and sometimes we can also use them for general purpose for writing email,

creating the blogs, creating the documentation. So we can do many things

documentation. So we can do many things by using these LLMs right and we can use these LMS with the general general language. So we don't need to learn any

language. So we don't need to learn any programming nothing just basic English knowledge is enough to work with these LLMs. The most important thing is prompting this is very important. So

even though if you know the language if you're not providing proper prompt you will you will not get the proper output or proper response from the LLM. So you

should also know how to write a proper prompt in English language. So the

prompting is also very important. So

what is a prompting? Prompting is a a mechanism to provide the instructions to the LLM. So prompting is basically what

the LLM. So prompting is basically what providing providing inputs providing inputs or providing

instructions to the LLM. So prompting is basically providing inputs or instructions to the LLM. So we have to clearly provide the proper instructions

to the LLM. So then you will able to get the accurate output from the LLM. Okay.

So the LLMs basically what it will do based on the data which we already trained. So based on the data it will

trained. So based on the data it will able to generate some new content. Okay.

But it cannot act.

It has certain brain because we can instruct some data right. So it it can learn but it cannot act. So there are two things actually learning and acting.

So LLMs can learn the things but it cannot act that cannot execute. Okay. So

LLM can answer your questions. So if you ask something it can explain your topic.

And suppose if you ask LLM to generate some piece of code it can generate the code. It can generate the automation

code. It can generate the automation script but it cannot execute the script.

It can just only the generate the script based upon your instructions but it cannot execute it first of all.

Similarly, it can create an email. It

can generate the email message but it cannot send the email. So there is a two different things learning and acting. So

the main advantage of LLMs are what they can learn and then you can give the response but what is the limitation of

the LLMs? they cannot act or they cannot

the LLMs? they cannot act or they cannot uh do any activity or they cannot perform any action on the real time. So

that is a major limitation for LLMs. So for example, if I use charg and if you ask to write some test case, it is able to create the test case based upon your requirement. If I just provide simple

requirement. If I just provide simple requirement in the for in the form of prompting and it can generate the test case but it cannot execute the test case on the real browser. If I ask charg to

execute the test case, it cannot right.

So it can have it it has a brain. It can

learn things. It can generate the content new content but it cannot act.

It cannot execute. It cannot perform any action on the runtime. So that is the main limitation of this LLMs. So there exactly the agent come into picture. So

because of this limitations agent come into picture. So now what is an agent?

into picture. So now what is an agent?

So what is the limitation of LLMs? LLMs

can think, can learn, can provide the response, they can generate the output but they cannot execute. They cannot run anything. They cannot run your test

anything. They cannot run your test cases in the real browsers. They are all limitation from LLMs. And then agent come into picture. So what is the agent will do? Agent will actually perform the

will do? Agent will actually perform the actions. Agent can actually perform the

actions. Agent can actually perform the actions. agent can actually execute our

actions. agent can actually execute our automation test scripts on the real browser. It can interact with your

browser. It can interact with your browser. It can interact with your

browser. It can interact with your databases. It can interact with the

databases. It can interact with the backend APIs. It can access all the

backend APIs. It can access all the resources which are you have used at the back end even though they are remotely available or they are available in your local system also. So for example if you

are using charg don't have knowledge of about your files your computer files charg don't have any knowledge about the softwares which you have installed on your local system

right so but agent will know that so agent can execute so agent will able to access all the resources local resources or reot resources accordingly whatever

instructions you provided accordingly the agent will able to execute them on the real browser or real data databases or APS wherever you want. So we can say

a a uh agent is a basically an assistant.

A uh we can say it is an assistant that actually does what the LLM says. Okay.

So it is an assistant. Agent is an assistant that actually does what the LLM says. It can perform the actions.

LLM says. It can perform the actions.

It can perform actions or we can perform execution at the real time. So agent

alone cannot do any action. So we have to again provide some instructions from where we have to provide the instructions from the LLM only. Okay. We

will integrate LLM one of these LLM along with the agent. So LLM if I use only LLMs we cannot perform any action or execution on the real time. We can

simply generate the data or we can generate the automation scripts but we cannot execute them on the real browsers on the databases. So LLMs cannot access them all the resources whereas agent

will able to access everything all the resources browser databases APIs. So

wherever you have given the access to the agent agent will able to access everything but to perform action we have to again provide the prompt through LLM

only. So whatever LLM is provided based

only. So whatever LLM is provided based on the instructions the agent will act agent will perform the action on the real time. Okay. So based on the

real time. Okay. So based on the limitation of this LLM agents are developed.

But now agent and LLM how we can integrate these two. So agent is having LLM is having the brain. Okay that's

fine. agent can do the action but there should be something which should provide the instructions or protocol right so

there exactly MCP come into picture model context to protocol model context protocol so let's try to

understand this llm can generate the content it can just think and provide the response it cannot act agent can act

based on the instructions we provided through LLM. But even though if you want

through LLM. But even though if you want to if you want agent to be acted so we need to provide a proper pro protocols like for example agent can access the

agent can access the database and it can do something on the database. Agent can

access the browser, it can perform certain things on the browser. But what

it should do that we have to provide instruction like okay agent can open the browser but after opening the browser what we should do what agent should do we need to open the URL we need to interact with elements right we need to

provide all those instructions who will do the those all these things MCP will do that so agent will perform the actions based on a protocol right that

is provided by the MCP so MCP provided a protocol based on this model context protocol Agent will actor. Agent will

actor. So now what is the relation between these things? So what MCP can do? What else MC MCP can do? MCP can do

do? What else MC MCP can do? MCP can do browser automation. It can open the

browser automation. It can open the website. It can fill the forms. It can

website. It can fill the forms. It can click on the buttons also. Okay. So all

these things we can do through agent.

But to perform those things through agent again we have to provide the protocol through MCP. Okay. Through MCP

only it is possible. Only with agent we cannot do. So all three is interrelated.

cannot do. So all three is interrelated.

So in the first layer LLM come into picture. So what we can do with LLM? We

picture. So what we can do with LLM? We

can provide the prompt. So once it is provided the prompt then agent will perform the action. Agent will perform the action. On what basis agent perform

the action. On what basis agent perform perform the action? It can act on everything. It can access the resources,

everything. It can access the resources, databases, browsers and everything. But

we need to provide proper protocols to this through MCP. So here MCP come into picture. LLM agent and MCP. Okay. So

picture. LLM agent and MCP. Okay. So

basically LLM that understand the content and generates the new content. So what agent will do? Agent

content. So what agent will do? Agent

will assist to the LLM to perform the action. And

MCP works like a bridge between LLM and agent. MCP works between a like a bridge

agent. MCP works between a like a bridge between LLM and agent. So without MCP agent alone cannot do anything because even though agent is able to access all the resources

local or internal resources or external resources but it cannot perform any action until we provided a protocol or until provide certain instructions

through MCP. Okay. So this is the

through MCP. Okay. So this is the relation between the LLM agent and MCP.

Okay. So MCP is a protocol actually. So

which is provided to the agent and then agent will interact with everything.

So this is how the relation between LLM agent and MCP. So all LLMs comes under the generative AI. On top of that based on the limitations of LLMs agents have

been designed or developed and uh only with the agents we cannot achieve anything. So we should implement

achieve anything. So we should implement MCP. So there are multiple n number of

MCP. So there are multiple n number of MC MCPS are available in the market like we have a play rate MCP through which we can generate the playright automation code or automation script. We have a

selenium MCPS. We have a database

selenium MCPS. We have a database related MCPS. We have a ABA related

related MCPS. We have a ABA related MCPS. Different technologies are having

MCPS. Different technologies are having different kind of MCPS. So if you have such type of MCP we can work with that particular kind of technologies. For

example, if you want to work with the play rate automation, we need a playright MCP. Similar if you want to

playright MCP. Similar if you want to work with the selenium automation then we need a selenium MCP and suppose if you have a database designer or a database developer you want to access the database then you want to do certain

transactions or if you want to implement some database scripts then we need a database related MCPS. So in every area the different type of MCPS are developed which are currently available in the

market but it is ongoing process still there are so many MCPS coming into market. So this is the main thing. Okay.

market. So this is the main thing. Okay.

So LLMs are already available in the market. There are n number of LLMs are

market. There are n number of LLMs are there. We can we already know about

there. We can we already know about charges.

These are the popular LLMs. There are n number of LLM still there in the market and they are coming into market and agent and MCP. Okay. So this is relation

LLM, agent and MCP. So just understand LLM is basically can think and generate the new content. It cannot perform any action. Agent can perform the action. It

action. Agent can perform the action. It

can do the actual execution through browsers or database APIs everything will able to access but to perform such actions we again should provide certain

protocols or instructions through MCP until unless we have an MCP agent cannot work okay so that is all about so for example simple example I'll tell you so

agent and MCP suppose there is a person and if if you ask the person should go somewhere right that That is instruction actually but

until unless you set where he should go right if you if you don't tell the destination actual destination he won't go anywhere so that is the reason agent can do that agent can perform the

activity but MCP should provide the protocol to the agent and then only it will perform the action or execution on the real browsers or databases or APIs

or any other resources and LLMs cannot access our local resources like our databases our automation script Nothing will be accessed by LLM but agent will

able to access and MCP will able to provide the protocols to the agent and according it will perform the actions on the browsers. Okay. So this is a

the browsers. Okay. So this is a relation between LLM agent and MCP.

Okay. So everybody's clear so far. So

these are the three things we need to first understand then I'll show you practically how we can use LLM agent and then MCP.

Okay.

Now, so the next thing is I'll try to introduce. So we are not going to

introduce. So we are not going to directly work with this LLMs chart GBT Google because everybody knows about LLMs how we can use right simply every

day almost everyone uh using these LLMs on day-to-day basis for a different purposes. We already know about it. But

purposes. We already know about it. But

here the thing is how we can utilize them for our automation. Okay, how we can make our tasks faster through your automation. So let us see how we can do

automation. So let us see how we can do this. So before that we have something

this. So before that we have something called GitHub copilot. This is a very beautiful tool GitHub copilot.

So this is actually uh kind of LLM. This is also comes under the LLM GitHub copilot but especially designed for coders. So whoever is developing the code or whoever is

writing the automation test scripts right especially for those type of people the GitHub coal is GitHub copilot is got designed this is also a kind of a

framework okay so GitHub copilot is especially designed by GitHub itself and what exactly the GitHub copilot will do so whenever you start writing some

automation code right so before writing the entire code you can get the suggestions autom automatic suggestions by the GitHub copilot. So GitHub copilot

will provide you automatic suggestions when you're writing your code or when you're implementing some piece of code.

But from where this GitHub copilot will get the code automatic suggestions from the existing GitHub repositories. So in

the GitHub remote repository there are millions of repositories are created by millions of people. Right? So there are n number of repositories are there in our GitHub repositories. So what GitHub

copilot will do when I integrate this GitHub copilot in your automation or in your IDE whenever you start writing the code it will automatically suggest everything sometimes it will also

complete your code without writing the entire thing GitHub Collab will automatically give the suggestions and also it will automatically complete the code and sometimes if the code is having

some issues or some errors GitHub copilot will automatically rectify those errors. So that is also can be done by

errors. So that is also can be done by the GitHub copilot and also if I ask any query GitHub copilot will give the response because this is also part of

generative way. It is also an LLM. So

generative way. It is also an LLM. So

there are many uses are there. So is

this is especially designed for coders and again GitHub copilot is also internal use either charg so multiple LLMs sometimes we can use we

can choose actually in the GitHub co pilot it is supported multiple LLMs and we can use one of these LLM so GitHub copilot is a kind of a framework which

is implemented by taking help of LLM along with the GitHub repositories and especially designed for what designed for coders

means what? Developers or automation

means what? Developers or automation testers. Okay, designed for coders or

testers. Okay, designed for coders or automation testers. GitHub copilot is

automation testers. GitHub copilot is available. Okay, so now we'll see how we

available. Okay, so now we'll see how we can install GitHub copilot. And this

GitHub copilot supported almost in every IDE. We can install it with Visual

IDE. We can install it with Visual Studio Code. You can install with

Studio Code. You can install with PyCharm and uh all almost every kind of language, every technology supported nowadays. So, GitHub copilot is

nowadays. So, GitHub copilot is available for all kinds of ids, all kinds of languages.

Okay. So, everything is free. So, GitHub

copilot is totally open source. Okay.

So, we'll see how we can use this GitHub copilot along with our PyCharm. Okay.

There are multiple usages are there. So,

I will show you one by one. So, how we can use this GitHub copilot and how we can use it. So first of all let us see the installation part. How we can install GitHub copilot in your Pycharm

IDE. Okay. So let's go to PyCharm. This

IDE. Okay. So let's go to PyCharm. This

is our PyCharm. So to install GitHub copilot the first important thing is you need to have GitHub account. I hope

everybody is having GitHub account.

GitHub login. That is the first thing we need to have. And the second thing is uh yeah that is the only thing we need to have first. Okay. GitHub account is

have first. Okay. GitHub account is needed. So how we can install GitHub

needed. So how we can install GitHub copilot from IDE? Okay, simple. You need

to go to settings here. Go to file and go to settings and inside the settings you can see something called uh plugins available. Right here is a plugins.

available. Right here is a plugins.

Inside the plugins you can see marketplace. You need to go to

marketplace. You need to go to marketplace. So here you need to search

marketplace. So here you need to search for GitHub copilot. Simply say GitHub and you can say copilot.

So once you've searched for GitHub copilot here it will show you GitHub copilot option. So you just look at the

copilot option. So you just look at the icon. Okay. So there are many things you

icon. Okay. So there are many things you can see but this is the main icon you can see. You need to select this and or

can see. You need to select this and or you can see the install button like this because I already installed this copilot earlier. I'm not able to see the install

earlier. I'm not able to see the install button. But in your case first time when

button. But in your case first time when you installing this GitHub copilot you will see install button here. Okay. Just

click on install button. This plug-in

will be installed. And once the plug-in is got installed, it will ask you to restart your IDE. Then you have to restart your PyCharm. You need to restart your PyCharm. So that's the

first thing. So once you restart your uh

first thing. So once you restart your uh PyCharm, then your GitHub copilot is ready to use in your PyCharm. This is

one simple installation. Okay. And not

only installation, you have to enable this. You have to enable this. I'll show

this. You have to enable this. I'll show

you. So once you have installed this GitHub copilot, it will restart your PyCharm ID. So once your PyCharm ID is

PyCharm ID. So once your PyCharm ID is re uh restarted then again one more time go to settings and here under the tools

under the tools uh you can see copilot option here GitHub copilot option here.

So inside this GitHub copilot option there are multiple options you can see here. Okay. So if you are able to see

here. Okay. So if you are able to see this go GitHub copilot that means you have already installed this GitHub copilot plug-in. Okay. This is the step

copilot plug-in. Okay. This is the step number one. So once you have installed

number one. So once you have installed GitHub copilot then the next thing is what we need to connect. So you will able to see one icon in the uh right

side in the bottom section. You can see currently this is not signed. So if I click on here you will see these options.

Okay. First time. So when you open the GitHub copilot in PyCharm IDE you will see these options. So now what we need to do next step once we have installed

plug-in we need to log into the GitHub with your account. So whatever GitHub account you already have with that account we should log into the GitHub because GitHub copilot works based on

the GitHub repositories. So even though it is getting the code from any number of repositories but you need to authenticate with your GitHub repository account. Okay. So once you install this

account. Okay. So once you install this GitHub copilot plugin then you'll get this small icon here and then you need to just click on log to GitHub. So once

you click on this login to GitHub it will show you the device code. Okay. So

once you copy this just copy this device code. Once you copy this click on copy

code. Once you copy this click on copy and open and then it will automatically open your GitHub account. So, GitHub URL. It will

GitHub account. So, GitHub URL. It will

navigate to the GitHub URL. So, whatever

account you already have, you need to sign in with this account. Sign in. And

this is account I'm using. And if you have another GitHub account, you need to use a different account. So, simply

click on continue. And then whatever code you copied in the previous step, just you need to copy this code. Click

on continue. So, once you have done it, now it will show you something. Full

control of private repositories. Update

GitHub action workflows. By default,

both the options will be selected and just click on authorize GitHub.

It will show you something. This

authorization were requested from so and so place and so and so date. Make sure

you trust this device as it will get access to your account. So once you get this, you simply say authorize GitHub.

So once you click on authorize GitHub, then it will be authorized. So let me share a small document because I already done this multiple times. You can see this. This is the screen just now have

this. This is the screen just now have seen. Okay. So just click on continue

seen. Okay. So just click on continue and then it will show you the code and then once you authorized right. So you

will see this one authorize GitHub copilot ID plug-in. So simply click on authorize GitHub. So once you have done

authorize GitHub. So once you have done it will ask you the password and simply simply say confirm. So once you have confirmed then you'll get a congratulations message. So this is how

congratulations message. So this is how we can simply log into your GitHub repository. So after that you will able

repository. So after that you will able to see all the other options in this icon. So I'll show you quickly now.

icon. So I'll show you quickly now.

Yeah. So it is enabled my button. So I'm

just clicking authorize GitHub. So once

you click on this button, it will ask you uh about my GitHub password. So let

me provide my password.

Okay. So I'm just providing my GitHub password.

So once you provide it, click on continue or confirm. Now congratulations

you have set. So once you come back to the your IDE and it is giving some the login to GitHub failed reason. Okay, it

is giving some error. Let me just close it and because it is open lately. So that's

the reason it's giving some error unknown error. So let's try one more

unknown error. So let's try one more time. And I'm just closing once again.

time. And I'm just closing once again.

And if you're facing some issues at the time of login, just retry once again.

Okay. So I'm just again trying to login.

So log to GitHub and again copy this code and copy and open.

Okay. So just simply say continue.

Copy this code. Again click on continue and click on authorize GitHub.

Congratulations. So once it is done again you need to go back and it is taking few seconds. Now you can see successfully logged into GitHub for GitHub copilot. So you should see this

GitHub copilot. So you should see this message. So once you are able to log

message. So once you are able to log into the GitHub now we can see green color tick mark and if you open this icon you will see some options here. You

can see open chart edit settings. There

are many settings view kota usage there are many settings are there. Okay. So

this is the second step. First we have to install GitHub copilot plug-in through uh settings and this plug-in option and the second thing is once you get an icon

you need to log into the GitHub with your own account. Okay, these are the two things we need to do. So once you successfully log in now your GitHub

account is ready. Your copilot is ready to use. Your copilot is ready to use. So

to use. Your copilot is ready to use. So

if you want to check the settings once again you can just go to the file and go to the settings and you can see the GitHub copilot this option you can see

right so these are all the options you are able to see here so now once we have installed GitHub copilot now let us see how we can use in our automation so what

we can achieve with this GitHub copilot so I will provide the documentation also so step-by-step documentation so you will able to install this very simply

So once we install GitHub copilot, so the next thing is how we can use GitHub copilot. So there are multiple uses we

copilot. So there are multiple uses we have. So we can use it for code

have. So we can use it for code completion. We can ask question and

completion. We can ask question and answers in the form of comments. And

suppose if you already have some piece of code, so you have not understood that part code. So then you can ask copilot

part code. So then you can ask copilot to explain that part code and sometimes you want to simplify the code or you want to refactor the code. So we can also simplify the code and also it is

provided the chart window whatever you want you can ask and we can generate the documentation and we can put the window chart window wherever you want. So

currently this chart window is provided here right suppose if I open the chat box here currently the chat box is based here it is displayed but if you want to move this chart box wherever you want you can move it. So I can just click on

this option three dots and you can see this view mode and uh here you can see something called window. I click on this window. Now the chart box you can put

window. Now the chart box you can put wherever you want. Okay. And again if you want to pin it you can just go with view window view mode and then simply

say doc pinned can use this option. So

this is how we can open the chart box from this GitHub. So now let us see how we can use this uh GitHub copilot in the different ways. So we can use code

different ways. So we can use code completion question answers in the form of comments we can provide. we can ask to explain the code. We can simplify this and many more things we can do.

I'll show you one by one. I'll

demonstrate. Look at this carefully.

Okay. So the first thing code completion. So what is code completion

completion. So what is code completion feature?

So I will also write here. Okay. So the

features of GitHub copilot.

So, GitHub copilot.

So, the first feature is code completion.

Okay, GitHub code pilot you can use for code completion. So, what is code

code completion. So, what is code completion? I'll show you. So, let me

completion? I'll show you. So, let me just create a new directory for grace class. I say a testing. Okay. So, we

class. I say a testing. Okay. So, we

already enabled GitHub copilot and we already connected to GitHub repository through copilot which is available. Now

what we can do simply create a new Python file. I call it something test

Python file. I call it something test copilot demo. Okay. I'm just simple

demo. Okay. I'm just simple giving a simple file copilot demo. Okay.

So what is code completion? So whenever

you start writing the code, it will automatically provide the code automatically. So for example, let's say

automatically. So for example, let's say I just want to write a playright test. I

want to write a playright test. So what

you will do initially we will import required package right? So see how the code completion works. So I'm just trying to import something you can say from

from playrite dot the sync API from that I'm trying to import

something called playright and something so let it be there. Okay now I'm starting simple writing a code. Now we

can see this as soon as you start writing simple keyword it is automatically suggesting I'm writing simple play rate code play rate test so def let's say test

login see this this is the beauty as soon as I just type test login it is automatically identify what I'm going to write and it is pull entire code from some gith

repositories and it is giving some auto suggestions look at this all this code is automatically provided by copilot But sometimes it is relevant or

sometimes it may not be relevant. If it

is relevant, you can simply put tab and the code is automatically available. And

if it is not relevant, you can just continue typing something. But for now, I don't I I don't have any relevant for this. Suppose let me remove this. I'm

this. Suppose let me remove this. I'm

not this content is not relevant for me.

So what I'm going to do is I'm just continue to do some piece of code. Let's

say test login. Here I'm going to create a page picture for example page should come from the page.

Now we need to import this. Now you can see as soon as you enter it is started giving all the examples. Now I'll import this page.

See as soon as you type enter here it is automatically giving a code suggestions. This is

called code completion. Okay. Simply you

can if you want to have this piece of code you can simply put tab or if you don't want you can just continue to this for example I don't want this code still continue to type so I'm saying page dot

go to and here whatever URL you want you can just simply provide the URL here simply I can say https

https from each and every step it will be very very helpful ww google.com so I just type one statement Now after that

enter. So again it is giving assertion

enter. So again it is giving assertion suggestion also. Suppose I want to have

suggestion also. Suppose I want to have this statement. Simply say tab. Now

this statement. Simply say tab. Now

another step is suggesting. What is

suggesting? It is searching for some search box and it is typing some playright. Okay. Now I want to have this

playright. Okay. Now I want to have this step. Again click on the tab. Simply

step. Again click on the tab. Simply

entering another one page. Press. It is

pressing an enter key. Now again I want to have this. Press enter. So if you want to have that piece of code, you can have it by pressing enter. And if you do not if you do not want to have it, you

can simply continue to type.

Okay. So this is how it will automatically suggest it will automatically identify what you are going to type based upon the previous knowledge. So

because in the repository there is millions of millions of code is already available. So based on that content it

available. So based on that content it is guessing what exactly you're going to write and accordingly it will start giving the suggestions. So this is

basically called code completion. The

feature of code completion from g uh github copilot.

So if it is giving you different suggestions then you can reask again. So

you can also provide the comment. I will

show you that. Okay. So this is code completion future. So code completion

completion future. So code completion feature means what? You can simply start typing accordingly. It will give the

typing accordingly. It will give the suggestions. If you want to have the

suggestions. If you want to have the suggestion, press tab key. And if you do not want to have the suggestion, you can just continue with that. Type continue

this. Okay. So this is simply where we can use copilot for code completion.

Okay. So similarly suppose you have written something here, right? Suppose I have written some piece

right? Suppose I have written some piece of code just two lines of code I have written.

Suppose I want to ask like suppose what exactly this code is doing right suppose if I not understood something okay let me just copy some piece of code from our

previous examples okay so let me yeah so get this piece of code I'm just for sample I'm just taking this

code okay and let me import this okay so this is a piece of code I have written in my code. So now I want to understand this code because I don't

know about this code. I just want to ask what exactly it is. So then what we can do is we can just select this code or suppose if you not understood a specific

line from this code. Okay. Then what you can do whichever line you have not understood you can select that line or if you not understood the entire code you can select the entire code and

simply right click here and you can just go here GitHub copilot. Here there are multiple options. You can see explain

multiple options. You can see explain this. You can see this option here and

this. You can see this option here and also you can see this option in the icon also. You can just open this icon and

also. You can just open this icon and open the chart and here you should able to see the options.

Okay. So it is not visible here and you can just directly right click on this code and right click and you can go here GitHub copilot and explain this. So once

you select this option it will explain the entire code whatever you have written. See each and every step it is

written. See each and every step it is clearly explained in this. So what

exactly we're doing in this step and what we are doing in this what we are doing in this. So each step is clearly explained here.

Okay. And this is one feature explaining this. Another feature suppose uh if

this. Another feature suppose uh if there is any uh issue or error in this particular code, we can easily fix it.

Suppose here uh I'm getting my URL equal to page URL. Page dot URL. I put bracket here. I put bracket here. And uh let us

here. I put bracket here. And uh let us see whether it is working or not. Let me

just run it first.

I just ran my code. So by default it is running on headed headers mode. So now

it is got failed because I put this uh bracket. Okay. So because of that the

bracket. Okay. So because of that the string object. So because of this the

string object. So because of this the test is got failed. So now I want to fix this through GitHub copilot. So how we can fix this? You can simply select the code or if you select the specific line

also if you already aware of where exactly the error comes you can select the particular line or you can select the entire code and again right click go to GitHub copilot and simply say fix

this. So once you fix this it is trying

this. So once you fix this it is trying to identify where exactly the errors are there and it is trying to fix it. Now

you can see it is it found the error here and it is rectified here and this is the final code which is given. So if

you want to have this code here or you can manually modify this by seeing this rectification or you can directly have this code copied into your main window.

You can do that. So for example here this is enter code newly generated right. So you can see one option here

right. So you can see one option here copy code block or the second option is also the insert code block at cursor. So

one is for copying this enter code. Once

you copy we need to paste it over here.

Other option is we have something called insert code block at cursor. So wherever

you put cursor, this entire code will be inserted. So for example, I want to

inserted. So for example, I want to insert this code here. Wherever my

cursor is waiting. Simply click on this insert code block at cursor. This code

will be automatically inserted. So the

other code can simply remove this.

Okay. So we can do like this. So this is how we can also fix the code. Wherever

the issue is there, we can simply fix the code. So for example, let's say I'm

the code. So for example, let's say I'm just removing the bracket here. I remove

the bracket here. Copilot itself is suggesting here. When you just say my

suggesting here. When you just say my URL, the bracket is suggesting by copilot. Suppose if you forgot to put

copilot. Suppose if you forgot to put somewhere and then again one more time you can ask to fix it. You can just select the code, right click GitHub copilot

and simply say fix this.

Okay. Now you can see it is successfully fixed. And again you want this code

fixed. And again you want this code simply click on this insert code block at cursor and the new code is got inserted here and remove this existing one.

Okay.

So this is how we can see. So control

alt and everything will be properly aligned. So these are the few uh

aligned. So these are the few uh important features which we have for copilot. And I'll show you few more

copilot. And I'll show you few more features and let me just minimize it now.

So what is another feature? Suppose you

can also ask anything through comments.

Okay. So for example, let's say I have not uh written this. Okay. Now I have just opened my page but I don't know how to capture the URL of the page. Let us say

I don't know how to capture the URL of the page. So then you can ask copilot.

the page. So then you can ask copilot.

How we can ask? Put comment.

Okay. And here you can ask how you can ask you can put Q or question because it will understand better. Okay. So here

you can ask see it is automatically giving the session get the URL of the application. So you can simply write get

application. So you can simply write get the URL of the application.

This is also auto completion. So as soon as you put the comment and enter see this it is giving the statement. So as

soon as you put enter it is giving. And

if you want to have that statement, you can simply put tab and keep with the statement. So sometimes when you're

statement. So sometimes when you're writing the code also if you forgot or if you don't know how we can extract the URL or anything it can be anything if you if you're not able to proceed

further struck somewhere then accordingly you can provide a proper prompt in the form of comment and it will suggest you okay then if you want to have the

session keep it in your code by pressing tab key and if you do not want again you can ask suppose if you do not like this again one more time you can ask. So

that's how we can use copilot.

Okay. So it will fix the logical errors also. It will fix the syntax errors as

also. It will fix the syntax errors as well as logical errors. But the only main thing is you need to provide the proper prompt. So you can also provide

proper prompt. So you can also provide the prompt. So for example let's say you

the prompt. So for example let's say you select the code. Okay. And I can just open the copilot here. There is a open chat window. And in this window also you

chat window. And in this window also you can provide the prompt. So whatever you want to get the details or information if you want to fix it or if you want to refactor it simplify this anything you

can ask by providing the prompt in this chat box. So we no need to go to

chat box. So we no need to go to external tool like charg or gemini we no need to go and login there. So directly

with which is a built-in tool copilot is built-in tool which is available for every ID. So without going to the

every ID. So without going to the another window without going for chargit or any external LLMs we can directly use this copilot inside the ID itself. So

whatever you want to have you whatever you want to know about this code so you can simply provide the prompt here and it will give the details about this. It

can also fix all issues errors, syntax errors, logical errors, anything it will automatically fix. Okay, this is about

automatically fix. Okay, this is about copilot.

Okay, and one more thing so you need to understand one limitation of this. So

what is this copilot is able to do here.

It is giving the code autocomp completion means it is generating it and it is trying to fix the issues. It is

trying to simplify this. Suppose I just want to make more simplify this code.

Then simply select it, right click GitHub copilot and simply say simplify this and it will give you more simplified code just three lines of code which is given. Instead of writing four

lines, three lines you can simply insert. Now I can change within three

insert. Now I can change within three lines everything is done. So instead of writing a separate line for page URL it is put inside one single line. So

anything any code like you have a thousand lines of code 100 lines of code doesn't matter just select the code you can fix it the whatever errors you have it can be syntax errors or logical

errors or if you want to simplify this you can simplify this by using copilot.

So this is how we can use it. So you

need to understand the limitation of this. We are able to generate something

this. We are able to generate something new. We are able to fix the existing

new. We are able to fix the existing issues but we are not able to execute.

So our c-ilot cannot execute this code.

Our c-ilot will able to generate the code. Our c-ilot will able to fix the

code. Our c-ilot will able to fix the code. It can give the auto suggestions

code. It can give the auto suggestions completion everything but it cannot run this code at the real browsers. Why?

Because the co-pilot is not aware of our browsers. Copilot is not aware of our

browsers. Copilot is not aware of our databases or any resources. So that is a limitation of this LLMs or c-ilot. Even

copilot also internally using LLMs. If I look at here, these are the different LLMs configured along with the copilot itself.

Okay, copilot is just a wrapper or a tool which is developed by on top of LLM. So these are current LLMs which is

LLM. So these are current LLMs which is supported in this copilot. So what is the limitation of this LLM or copilot is it cannot execute it can just generate

the new content. It can fix whatever issues we have in the code or it can refactor your code. It can simplify your code and it can ask whatever you want by

just putting some comment here. Okay,

that's it. So that will automatically do. So even debugging you don't need to

do. So even debugging you don't need to do anything manually. So it will automatically go through everything and it will debug. Again the debugging is part of execution, right? That is also

limitation in copilot. So that we will able to achieve with MCP that I will come to that. Okay. So debugging and execution this is part of execution action. So that we cannot do through

action. So that we cannot do through copilot but c-allet will try to fix the issues with the existing code whatever we have written here and copilot will refactor the code. Copilot will simplify

the code but copilot cannot debug it.

Debugging is different. Fixing the

issues is different. Okay. If you have a syntax error logical error which will able to fix it. So debugging in the sense we need to interact with multiple files. Okay. Maybe here this is a file

files. Okay. Maybe here this is a file doesn't have any dependency on any other file but sometimes this is depending on some other file right so there is a

dependency so debugging and execution these are all part of action execution so that is taken care by MCP okay so I

hope everybody's understood about GitHub copilot so GitHub copilot once you have enabled once you connected to GitHub repository you will able to use all the feature from GitHub copilot And if you

do not want to use GitHub copilot, simply log out from here. If you log out from GitHub copilot, it won't give any suggestions.

Okay, it will still chart window is enabled. But you won't use it for your

enabled. But you won't use it for your coding purpose.

Okay, so simply uh you can use like this.

So just by writing comment also it will able to understand and it will show you everything.

So that's how we can use GitHub copilot.

So now we will go to next step. So so

far we have seen how to use GitHub copilot. The next thing is what is the

copilot. The next thing is what is the limitation of LLM? We are not able to execute. We are not able to run. We are

execute. We are not able to run. We are

not able to debug. Right? So now MCP come into picture. So we need to additionally integrate MCP along with

this ID. MCP along with this ID model

this ID. MCP along with this ID model context protocol.

As I said what is MCP? MCP is a kind of a framework which will provide the protocols to the agent which will provide the protocols to the agent. So

if I look at the copilot here and open the chat box here they already provided one option called agent. So whenever we are working with MCP instead of ask we

have to select agent here. Agent should

be there LLM should be selected then only MCP will start working. Okay. Now

we will go to next step. So how we can integrate PyCharm IDE copilot plus MCP. So all

three things should be there in one place. So then we will able to achieve

place. So then we will able to achieve we can do magic. Let me show you how we can install MCP server especially playright MCP server. So as I said there are multiple MC MCPS are available in

the market for different kind of technology. Suppose if you want to work

technology. Suppose if you want to work with the purely databases then you have to install the different type of MCPS.

Then if you want to work with selenium there is another type of MCP. For play

rate there is another type of MCPS available.

Okay. Anybody can develop their own MCPS nowadays. So Microsoft playrate is

nowadays. So Microsoft playrate is provided their own MCP for accessing playrate. But nowadays people are

playrate. But nowadays people are developing their own MCPS according to their requirement. But they are not more

their requirement. But they are not more secure. They are not authenticated. But

secure. They are not authenticated. But

the play rate MCP is more authenticated because it is provided from play rate itself Microsoft as a back end. So we

can 100% there is an assurance we can still use it for our automation. Okay.

Now let us go to the next level how we can use Pycharm IDE GitHub copilot along with playrite MCP. So first we need to

go with a few installation and I'll show you step-by-step process.

Okay. So now we'll see how we can integrate MCP along with the GitHub copilot and PyCharm. And with this combination we will try to generate our

test cases along with execution. So by

using copilad we are able to generate the test cases but we are not able to execute. But now we are integrating MCP

execute. But now we are integrating MCP also. So once we have installed or

also. So once we have installed or integrated with MCP we can generate web test cases or UI test cases even API test cases also we can generate automatically

and just we need to provide the manual test case. So the steps simple in

test case. So the steps simple in English language and our MCP will automatically generate the test cases by taking help from LLM and agent and not only generating the test cases it will

automatically execute the test cases and at the time of execution if there are any issues it will automatically fix those issues also.

Okay. So that is called as autohealing mechanism. So autohealing mechanism is

mechanism. So autohealing mechanism is in the sense what it will automatically heal the automation code. So there

exactly debugging come into picture. We

don't need to debug anything manually.

So automatically our MCP will automatically debug by applying autofilling mechanisms. And this process is basically called wipe coding. So by taking help of MCP we

wipe coding. So by taking help of MCP we are able to generate the automation scripts which is called as wipe coding and we are able to run the tests and automatically fix the issues which is

called autohealing feature. So these are all part of MCPS.

Okay. So let us see it will automatically identify all the elements on the web page everything because MCP will able to interact with the browser interact with the elements it will

access the DOM also all types of locators it will able to capture it okay I will show you the magic now so let us see how we can enhance this so what are the prerequisite we already have Python

so we need to install Python if you don't have and PyCharm ID we already have it and gith GitHub account also we have it and if you already installed copilate we need to log into the GitHub

and additionally we need NodeJS guys because playrate MCP we can install only through NodeJS okay so playright MCP we can install only through NodeJS so

NodeJS is basically a software which we need to install which we need to install otherwise we cannot use playright MCP okay so to install NodeJS that's a

prerequisite I'll guide you step by step you need to just download the software go to this location download this software playrate uh MSI file and if you are working on Mac you will see DMG

file. So once you downloaded just go to

file. So once you downloaded just go to the steps installation next next and your NodeJS will be executed. So once

the NodeJS will be uh installed the finally you need to just configure the path in Windows like you need to go to system properties go to path and here just you need to add NodeJS location

like program files NodeJS this is a default location and after that you can verify the version which is NodeJS properly installed or not. So you can just go to command prompt and this is a

prerequisite to work with the play rate MCP especially. So go to the command

MCP especially. So go to the command prompt and then you need to execute this command. I already installed let me show

command. I already installed let me show you node - version. So once you execute it now you

version. So once you execute it now you can see the version. Similarly npm node package manager will automatically comes along with the NodeJS because that is needed actually to install all other

packages on top of NodeJS like we have a Python and pip right. So once you install Python you will get a pip using pip you can install all other plugins.

Similarly, if you install node, you will get npm. Through this npm, we can

get npm. Through this npm, we can install other plugins. So, this is a prerequisite to work with the playright MCP especially because playright MCP is

developed based on NodeJS itself. So,

these are the prerequisites. Okay. So,

yeah, Python, NodeJS should be installed. ID we already have and GitHub

installed. ID we already have and GitHub account with a copilot. So, copilot we have to install. So, install and configure GitHub copilot in PyCharm. So

we already given this document just now we already completed this step. So we

need to go to file and settings and then we need to install GitHub copilot. So we

already done this step earlier and we need to log into the GitHub copilot through ID. So we already done all these

through ID. So we already done all these steps. Okay, you don't need to repeat

steps. Okay, you don't need to repeat again. So once you have done this GitHub

again. So once you have done this GitHub copilot PyCharm then this is the new step. We need to additionally add MCP

step. We need to additionally add MCP server and we need to enable it. So to

install this MCP server what you need to do is we need to just minimizing this copilot. Copilot should be there still.

copilot. Copilot should be there still.

Okay. And what we need to do we need to go to file settings again and go to the same plug-in section here.

So to install copilot also we have done in the same place. Now go to the marketplace and search for MCP

and you can see MCP server. So this is a plug-in. You can look at the icon. Okay,

plug-in. You can look at the icon. Okay,

there are multiple things are there. You

need to correctly install MCP server.

You need to install this because I already installed. So it is just giving

already installed. So it is just giving a check box here. So you can see the button like this install. You just click on install. Once you install this MCP

on install. Once you install this MCP server, it will ask you to restart your PyCharm IDE. So once it is restarted,

PyCharm IDE. So once it is restarted, then your GitHub copilot is ready along with GitHub account login and MCP server is also ready. So once you have

installed MCP server once you restart your PyCharm ID every both of them you can see here. So you can just go to the tools menu. So here you can see MCP

tools menu. So here you can see MCP server. So you can see this copilot we

server. So you can see this copilot we already installed. Second thing is MCP

already installed. Second thing is MCP server. So just now we have installed.

server. So just now we have installed.

So these are the two components you should able to see under the tools and both the components we have to install only through plugins only. Here you can see the plugins plugins and the

marketplace. Okay. So the first thing is

marketplace. Okay. So the first thing is we need to install this MCP server. So

copilot is ready. MCP is ready. Now we

need to combine them because copilot is a different component. MCP server is a different component. So now we need to

different component. So now we need to interact with this. Right? Our copilot

should interact with the MCP server. So

both are two independent plugins. So now

we need to integrate them. So to

integrate them so we need to first go to file settings tools MPC MC MCP server.

We need to enable MCP server. So how we can do this? Again go to same go to the file and again go to the settings and

here go to MCP server under the tools.

Go to the tools and here you can see MCP server. Here

you can see small check box enable MCP server. You just select this box and it will give you some kind of warning enable MCP server. Simply click

on enable. As soon as you enable it, it will show you few things here. You can

just ignore all these things. You don't

need to do anything with this. Simply

enable this. That's it. I say apply.

Okay, this is the one thing you need to do. And uh once MCP server is done and

do. And uh once MCP server is done and the next thing what we need to do is once it is enabled now again we need to go back to the model context protocol.

Go back to the model context protocol where it is sorry GitHub protocol. Uh

this is MCP server we already configured.

[clears throat] So there is something called there is one more thing I think in this GitHub copilot. Yeah

here you can see under GitHub copilot there is model context protocol. There

is one more MCP can see because we have just enabled here. So once you enable this MCP server then under GitHub copilot you can see this option model context protocol. So here what we need

context protocol. So here what we need to do is here we need to just click on this configure right. So in this by default you will

right. So in this by default you will see some piece of code automated code you should remove that part and place this one. So this is a code which is

this one. So this is a code which is provided by playright MCP itself. So

because if I just go to search for playright MCP okay just type playrite MCP and they provided their official document how to install this playright

MCP. We scroll down they have given this

MCP. We scroll down they have given this piece of code. This is the configuration code which we have to paste in the JSON file. So you need to just copy this code

file. So you need to just copy this code inside this MCP.json. So when you will get this file when you go to the file settings and when you go to the GitHub

copilot and model context protocol once you click on this configure here automatically MCP.json JSON file will be

automatically MCP.json JSON file will be open and in this file by default some piece of code will be displayed. You

have to remove that code and paste this is new code. Okay. So this partular code will be actually install the MCP client

MCP client to run the playright. Okay.

And this will actually having all playrate commands internally because how MCP will interact with the browsers? MCP

should know all the commands all the playright commands. So to know those

playright commands. So to know those commands we have to install this MCP client through this JSON so that MCP is aware of all play rate commands. So once

we have done it done so all our setup is done. So we can just simply close it. So

done. So we can just simply close it. So

now we have done installation. We have

done installation for copilot and we have installation for MCP. So everything

is done. So I put all the nodes here.

You can do step by step very carefully.

So first we have to install GitHub copilot and then we have to install MCP server from the marketplace and then we have to

enable this MCP server. After enabling

MCP server like this and then again we need to integrate MCP server with the copilot. So for that again we need to go

copilot. So for that again we need to go to tools and GitHub copilot model contest protocol. Click on this

contest protocol. Click on this configure and here there is one more option called allowed models that is optional actually. If I just go back to

optional actually. If I just go back to settings here, we already configured this piece of code, right? So apart from this, if you go to uh model config here

and here, allowed models. So currently

these are the models. Whatever models

LLM models you have selected, same models will be displayed in the copilot.

And if you do not want any other model, you can simply uncheck it. All models,

whichever model you want, you can use it. This is one more thing that is

it. This is one more thing that is optional actually. So here we can

optional actually. So here we can integrate MCP server with the copilot and once you configure the done. So once

you click on okay and this is the MCP.json file where you need to copy

MCP.json file where you need to copy this code and after doing all the things we need to restart our PyCharm ID once otherwise all these changes we won't be

affected. Okay. So let me restart my

affected. Okay. So let me restart my PyCharm ID. So let me close this exit

PyCharm ID. So let me close this exit and I'm just trying to restart.

Just wait for a few seconds. This is a play rate MCP documentation official documentation they have provided. So

play uh play rate MCP is know all the commands because as soon as you have installed with this JSON uh it is known all the commands all play rate commands.

So as of now this play rate MCP supported all these LLMs like cloud cursor goose and uh VS code all this the some of them are editor some of them are

LLMs current it is supported okay so it is still yeah now I have just launched my Pycharm MCP after installation of everything

okay so now we are ready to use MCP so now you will see real logic how MCP is able to generate our test cases based upon our prompt and how it is able to

execute our test cases how it is able to fix the issues everything I will show you practically now so now what we need to do is I will show you first how to generate the web UI test case okay I'll

also show you API test case now here I have designed one prompt see this this is a exact prompt you can modify according to your test case so but you can look at this what is the prompt I'm

giving generate and execute a playright test using pi test and playrate MCP.

So what I'm asking here generate and execute generate and execute both I have mentioned here and play rate test using

pi test and play rate mcp and this is my test scenario or test steps. So what I'm doing here just launch this URL search for t-shirts in the search box. I'm not

providing any locators nothing. It will

automatically find everything and then I'm verifying that butler t-shirts have been selected. This part title should be

been selected. This part title should be displayed in the search eraser. Simply

I'm giving three steps. You can try with multiple steps. And also I'm providing

multiple steps. And also I'm providing implementation requirements. What are

implementation requirements. What are those requirements? use pyest style

those requirements? use pyest style structure because pyest is having a separate kind of a structure to write the test cases like every test should start with the test underscore the function should start with the test

underscore so that's the reason I'm saying functions starting with the test underscore you have to clearly provide the prompt otherwise it won't be

generated properly then what I'm asking execute each step interactively through playright MCP tools at the time of generating the steps

It will ask you to interact. Okay. Do

you want to continue with this or not?

Then once all steps are succeeded, generate and save the file uh final test under scripts final test script as AI/

search product py. What I'm asking here once the test case is got generated once it is successfully executed then save the test case with this name under

AI folder. So here I already created one

AI folder. So here I already created one folder called AI testing or you can change it AI testing or AI folder wherever. So I'm trying to change it for

wherever. So I'm trying to change it for example let's say AI testing folder. So

whatever the tests have created, generated and executed, stored that part test inside this AA testing folder which I already created here. And what else?

Run the saved test automatically in headed mode.

That is up to you. Okay, if you don't want to use headed mode, you can also run headless mode. So I'm just specifically want to show you this.

That's the reason I'm specifically providing this instructions. Run the

saved test automatically in headed mode.

Browser should be visible. Okay. In any

step, if any step or assertion got failed, for example, at the runtime, identify and fix the issue, then rerun until the test passes. So, rerun also

should automatically taken care by MCP server. So, this part step will

server. So, this part step will automatically taking care of debugging, fixing the issues and rerun the tests.

So, we don't need any manual intervention here. So we need to

intervention here. So we need to properly provide the requirements and what expected output you want. A fully

working playright pyest test script and the test script should automate the complete flow described above and successfully validate the product name

in the results. This is a complete prompt. So accordingly you need to

prompt. So accordingly you need to design this and test case to test case little bit different but only steps will be changed but according to play rate python you no need to change the rest of

the structure because these step steps are almost same for all the test cases and later I will also show you how we can generate a test by using page object class page object model okay that is

also there so for now let's create test case let's try to generate the test case let us let it execute it and also After execution, it will also save the test

case inside the AA testing folder.

That's what I'm asking to MCP. Now, what

you need to do, simply copy this prompt.

Now, you see the magic here. Just copy

this prompt. Go back to your PyCharm IDE and open your GitHub. Open your G copilot. GitHub copilot. Open the chart.

copilot. GitHub copilot. Open the chart.

And here in this chart box, you have to select the agent. Okay? You have to select the agent and here whichever LLM you want you can select. So I'm trying

to use this GPT 4.1 which is latest and then whatever the prompt you have simply copy this prompt here.

Just a moment.

So I'm not able to copy directly from here. So I will put this inside the

here. So I will put this inside the notepad.

Okay. So now let us copy this paste it over here. So everything should be in the text format. Suppose if you have any file right this entire thing this entire prompt is saved in some text

file or some other you can attach that file directly. In that case you don't

file directly. In that case you don't need to write anything here. So whatever

is there in the attached file that will be executed by MCP. Now see the beauty you have to focus very carefully how it is going to do it. So I just providing

the prompt and simply say send just I have sent a prompt. Now see the beauty it is trying to understand.

It is trying to do everything and capturing everything. So you can see rand create file tool something is getting some warning. Let it be there.

It will automatically fix. Now you can see the beauty. It is automatically generating the code and also you can see we need to accept. Sometimes it will give you some notification. We need to accept then it will proceed.

Okay. Now it is asking to run your command. Test is got generated but it is

command. Test is got generated but it is asking us to run this. Simply say

continue.

Now it is started executing our test.

Collected one item. Now you can see it is opening the browser running the test.

So within seconds of time the test is got generated and execution is also completed successfully passed. And what

we have asked after successful completion save the file in aa testing folder you can see that search product.

So this is the file name which we have given in the prompt. So it is got generated file name. So this is a test which is got created. So again if you want to do some modifications in this

test we can simply do it. Okay. Now let

us try to use some other llm and see how it is going to generate. So what I will do it I can simply close this. I delete

this for now. I'm deleting the file.

Okay. Now let us give the same prompt one more time. So generate and execute play rate test using pi test and play mcp. Okay. Ex everything is same. Now

mcp. Okay. Ex everything is same. Now

I'm just copying the same thing. One

more time I'm providing but this time I'm trying to take another llm. So

whichever you want you can take. So

let's say GPT5 mini I'm taking. Let's

see now. and simply send. I think the previous one is still running actually.

So run the command in terminal edited mode. So we have still not completed

mode. So we have still not completed actually. Let me cancel this. Okay. Now

actually. Let me cancel this. Okay. Now

again everything we are starting from new.

Okay. So done.

Yeah. Now I'm giving the prompt one more time. Run it.

time. Run it.

If there is an error, it will automatically fix it and rerun it also.

So let it complete first.

Okay, it is working.

So, it is doing all the activities and in between it will ask you your permission also. Do you want to continue

permission also. Do you want to continue or not? Because that is also we provided

or not? Because that is also we provided in the prompt. So, it is generating the test. See depending on the network speed

test. See depending on the network speed it depends okay how faster it is.

Sometimes it will go very fast. Now you

can see how much time we will take to write this entire piece of code.

Lot of time right. So it is simply generated. So just accept and still it

generated. So just accept and still it is not completed. So it is asking run command internals. So what is this

command internals. So what is this command headed mode it is basically running the test. So simply click on continue. So it is started running the

continue. So it is started running the test.

See again it is open the browser.

past. Suppose if it is a failed, what will happen? We already provided the

will happen? We already provided the prompt here, right? So if any step or assertion fails, identify fix the issue, then rerun until the test passes because we already provided this instruction. If

anything is got failed at the time of generating the test. Sometimes there

will be some duplicate locators at the time test will fail first time. Even

though that is generated by MCP, test will fail. In that case, what it will

will fail. In that case, what it will do? It will automatically identify the

do? It will automatically identify the fix and then fix that issue and then it automatically rerun it. So we no need to rerun the test. So till the test passes

it will keep rerunning and keep doing some changes till the test is passes because of this prompt it will automatically take care of it. So once

it is done the new file is got automatically generated here and after that if you want to do some slight modification suppose if you want to convert this in

our own pyest style little bit this is also pyest format but it is giving a function format. So normally what you

function format. So normally what you will do you will pass a play playright fixture right here say playright fixture you will pass like this right. So we need to import this

this right. So we need to import this playright import this from sync API playright. So

like this and then you can just remove this part and this particular playright you can use install P. Now if you want to align this you can again simply right click

select this and right click and anyway we have a GitHub copilot everything right. So we can simply simplify this.

right. So we can simply simplify this.

Click on simplify this. It will make more simplify this code.

Clean up and refactor the code and simply can say again insert code block.

That's it.

And then close this or else you can simply select Ctrl Alt L. Everything is aligned.

L. Everything is aligned.

So this is how we can simply use GitHub copilot.

So are we doing any coding here? Simple.

We are not doing any coding. We are just frame the prompt properly. We have to just provide the instructions carefully.

Automatically the test is generated.

Execution is also happened. If there are any issues, it will automatically return. But currently we don't have any

return. But currently we don't have any issues in the code. But sometimes even MCPS also generate this code uh by mistake some taking incorrect kind of locators at the time test will fail and

automatically fix it one more time. Keep

on running until it passes. So this is the beauty of using MCP.

Okay. Now I'll show you one more example. This time we will generate the

example. This time we will generate the test case using page object class. But

as of now we are not using any page object class here. Right? So we will try to create a page object class and then create the test. Even page object class also created by MCP itself. So let me do

that. So let me give another prompt. So

that. So let me give another prompt. So

we already done this. Next one I'll show you.

same uh this time I will give you a different prompt okay uh with using page object classes we want to generate the test

okay so this is another prompt so what I'm doing here generate and execute play rate test using piest and play rate MCP same test scenarios got

changed so this time I'm opening another URL click on the login link in the top navigation bar enter in the username field. enter tested 1 23 in the password

field. enter tested 1 23 in the password field because we already aware of this application right so this is the one actually let me show you this so this is

the application so I'm trying to automate this so click on the login here provide our username and password click on the login and this is the flow I want

to automate and after that what you want to do verify that logout link is visible and also verify that welcome o appears at the top right of the page then click on the logout link verify that log login

login link is visible after logging out.

So these are all the steps I want to generate then implementation requirements are same and this time you want to generate page object class. So

apply the page object model design pattern for page structure and element management and save all page object files inside the project's AI directory.

So here we are using aa testing right a test same directory page object class also I want to store and follow the pi test style convention function starting with test execute each step

interactively using playright MCP tools to validate flow and elements once all steps are validated successfully then generate and save the file test script

as a testing and test login and log out this is the name of the test file and automatically execute the save test in headed mode and if any step or assertion

fails, analyze, fix and rerun the test until it passes. So this is complete prompt. Everything is mentioned. Now see

prompt. Everything is mentioned. Now see

the beauty how it is going to generate the page object class within seconds everything will be done. See same

procedure. So let me go to p copilot open the chat box and just provide the prompt as usual. Select the agent and whichever pilot you want to select.

Simply select whichever you want and simply send it. Now observe.

So it is taking all the steps prompt everything it is started doing. See

within page object classes got generated. Now test cases also got

generated. Now test cases also got generated within few seconds page object class test case everything is ready. If you

have a multiple page object classes even multiple page object classes also automatically create. I think you see

automatically create. I think you see multiple page object classes have been created.

Okay. Okay, now it will ask it is asking for execution. See login out page object

for execution. See login out page object class is created. Test case is also created. Now it is asking for execution.

created. Now it is asking for execution.

Simply say continue. Now it is started execution. Now you can see it is got

execution. Now you can see it is got failed somewhere. No module named AA

failed somewhere. No module named AA testing. Inside the AA testing

testing. Inside the AA testing everything is there but somehow it is giving an error. So what we should do now? It is giving some error. Right? So

now? It is giving some error. Right? So

again rerun continue because if there are any issues we are asked to rerun one more time right so that is the prompt is executing so again it is failed here that's the reason it is asking to rerun

run the command in terminal one more time so simply say continue we are asked to fix that issue and rerun again right so now it is started execution see this

so first time we got an issue and again it is automatically fixed and it is successfully executed so this time definitely it should class.

See it is still validating the log out link and the name which is displayed in the navigation bar.

So suppose again if it is failed again it will rerun one more time.

Yeah, now it is done. So somehow it is got failed once again. Uh okay again if it is failed again it will ask you to rerun one more time. So two times it is got failed. So again it is working on

got failed. So again it is working on it. It is working on fix.

it. It is working on fix.

So once it is working on fix it will give you the command again. Do you want to read? Now we can see it is identify

to read? Now we can see it is identify all the fixes right. So you can see

right. So you can see accept all and now again it is giving the same command to run. So now it is got fixed again. So again continue this. So until

again. So again continue this. So until

it fixes it keeps running the automation test. So now again it is open login log

test. So now again it is open login log out.

So I think now it will pass. So

successfully login.

It will take multiple reruns. So now it is almost successfully login. Somehow it

is g failed because of some time out page wait for selector. Right. So this

is 15,000 exceeded. Somehow it has got failed somewhere. Okay. So it is again

failed somewhere. Okay. So it is again still trying to fix it. Let it fix.

We'll wait for a few seconds.

Okay. So now it is trying to do some keep changes. Uh you can see it is

keep changes. Uh you can see it is actually this is the error. Uh it is giving actually uh the fix and if you want to accept this fix we need to click

on the tick mark actually. Okay. So we

need to just click on this tick mark and say accept right again continue this is asking changes. So once you accepted

those changes then it is trying to fix and rerun it. So this is the beauty of using MCP. So within fraction of seconds

using MCP. So within fraction of seconds everything is doing it. So if you want to do it manually creating page object classes creating the test case takes a lot of time.

Okay. So it takes multiple reruns or multiple times. So but this is the

multiple times. So but this is the procedure guys. I hope you understood.

procedure guys. I hope you understood.

So this is the procedure we need to uh generate the test cases and run the test cases. If there are any failures are

cases. If there are any failures are happening the test cases and in between if you want to do manual changes and still you can do those manual changes okay

according to the test but before doing all these things at least you have to aware of automation right so until unless you know the coding and implementation you cannot use this MCP

some play rate anything right so that's the reason first you have to learn everything you have to know how to write the automation code you have to learn how to create a page object class and then if you can start using MCP models.

So if I start using if even though if you face any issues or any issues you can easily fix them. So that is the main advantage of using MCP server. Okay. So

similarly different tools are having different type of MCP servers. Even Celm

is also having their own MCP server.

Celerium TypeScript if you're using Celium sorry if I using playright typescript combination again same playright server. So this is supported

playright server. So this is supported all three languages. So this playright MCP is implemented for playright javascript playright typescript playright java playright python all four

languages are supported with single playright MCP okay but selenium java selenium python is having a different kinds of mcps again they have implemented by some third party people

not by playr not by Microsoft so we cannot give guarantee on those mcps and security aspect okay so this is how we

can use MCPS uh to automatically generate the code automatically create the files automatically the execution will also automatically happens right so this will

saves a lot of time and this is a nowadays trend okay so this is a upgoing ongoing technologies in the current market so

this is a future so there are n number of MCPS are already came in the market still coming so you need to learn this and you need to upgrade This copilot

will support almost all type of ids and uh Eclipse I'm not sure but support VS code editor Pycharm it support intellig ID is also support and if you're working

with selium java try to use intellig ID which is having good number of features copilot supports right so MCP is also available for all kinds of technologies so there are different technologies

different MCPS are available if you want to work with the database side then database MCP is needed right So I'll show you quickly API also how we can generate API test case. So far we have

seen volley web test. So we can also generate API test. So what I can do is I will also give a new prompt

for API test.

So let's see how exactly API test will generate. So I'm just changing the

generate. So I'm just changing the prompt.

So generate and execute an API test using pyest and playright mcp. So what

is my test scenario? I want to set API endpoint. This is the API endpoint. So

endpoint. This is the API endpoint. So

if I run this API endpoint actually what will happen is it will return some JSON response. So this is our response. In

response. So this is our response. In

the response we have an ID title and some price description so many other things. So let us test this. So send the

things. So let us test this. So send the get request to the endpoint. Verify the

response status code 200. validate that

response. JSON contains the keys. So

these are the keys should be there in the JSON response and the data also verify the JSON response includes the following expected values and after that what I should do log the complete

response body to the console. I just

want to print in my console window and what are the implementation details?

Follow the Pyest structure. Execute each

step interactively using player MCP tools for validation. After successful

validation, save the generated script as again in AI testing folder with this name test API test MCP. If any step or assertion fails, again identify, fix the issues, then rerun until the test

passes. What is an expected output? A

passes. What is an expected output? A

complete and functional play rate plus pi test API test script that verifies the product details from the fake store API and logs the response body. This is

complete prompt. Let's see how it is going to generate API test case and it will run also automatically. Go back and go to copilot. Open the chat window.

Agent GDP 4.1 I'm selecting. Put the

prompt. Simply send it. So this time it is going to generate API test case within few seconds.

See this done.

Accept. And now it is giving the command to run it. We need to give the permission to run the test. So click on continue. Now it is started running my

continue. Now it is started running my test.

Done. You can see one test is got passed. So as soon as it get passed, it

passed. So as soon as it get passed, it is also saved the test case inside my AA testing folder. You can look at

testing folder. You can look at somewhere you can see this test API test_mcp.

So this is test which is got generated.

This is the API test case.

Okay. So you can use this MCP for API testing also.

So that again you need to include in the prompt. Okay. So how you have designed

prompt. Okay. So how you have designed the prompt depends on that it will behave. If you want to capture the

behave. If you want to capture the screenshot that should be also included in the prompt properly right if there are any failure just you need to capture the screenshot you need to add that point currently I have not added

anything in this prompt I just given plain prompt okay but that is anyway we are uh handling through framework right capturing the screenshot and everything so that is not needed to put in the prompt separately but if you are

creating individual test case without using any framework then you can add that in the prompter but anyway we don't do individual test case like this Right, we are going to follow the framework concept. In the framework, we already

concept. In the framework, we already have pyest.ini, conf test. py. So all

have pyest.ini, conf test. py. So all

the configuration files we have already handled how to capture screenshot, videos, trays and everything. So better

to don't include them inside the prompt.

So just write a clean prompt and keeping everything in the proper prompt and then it will do it. So it did not log in the console window. It should log actually.

console window. It should log actually.

It did not log or else we need to rerun once again.

Okay. So complete functional what we expecting if any so where we provided the prompt. So

verify the JSON response we have okay log the complete response body in the actually this is the console what we are calling here. This is yeah console or we

calling here. This is yeah console or we can just try to change the prompt log or print the complete response

body to the console as an output. Okay. So just need to change the prompt properly or if you're not satisfied with the output then you

can rerun it. No problem at all. Okay.

So you can just remove the code whatever you have created. Simply provide the prompt one more time.

and see this this time we'll see whether it is printing the output or not.

So everything is depends on prompt only guys. Okay. If you properly give the

guys. Okay. If you properly give the prompt accordingly it will do it. Now it

is asking to execute. Say continue.

So if I look at the test it is not added any step to print the response. See

here. Yeah, actually here it is printing response dot in the text format it is printing actually but it is not text right so it should be

JSON right it should be JSON this actually return the response so again you can rerun it so

and you can just provide only specific prompt run the test again.

So the same test again it is asking for command. Click on continue. We just do

command. Click on continue. We just do some modification right.

Okay. So my plan is you have reached the monthly chart messages. Go to upgrade to copel. So 30 days uh free available

copel. So 30 days uh free available guys. After that it will be we need to

guys. After that it will be we need to upgrade this or else we need to uh wait for another 24 hours something like this then you can use this copilot chart box clearly

giving you have reached the monthly chart messages let us try to use another GBT and see okay let me

rerun the test one Okay, I'm just trying to change the LLM and we'll see. No, this is coming from

we'll see. No, this is coming from actually copilot. Okay, so we need to

actually copilot. Okay, so we need to upgrade the plan. So this is a free version is provided but once the free trial is over, we need to again upgrade with this copilot or else you need to

create another GitHub account. So you

can create multiple GitHub accounts so that we can use another 30 days like that. But in the real time actually they

that. But in the real time actually they will uh upgrade it by paying something.

It is not totally free actually but the free trial is available 30 days we can use it. So that's it. My limit is over

use it. So that's it. My limit is over here. So I I hope everybody's understood

here. So I I hope everybody's understood so far right how we can use and how we can generate the test cases using MCP along with the play rate. So I'll share this document keep it with you. I

provided all the steps one by one how to install copilot and MCP server then integrate copilot with MCP server in PyCharm ID provided this command in

MCP.json JSON and this prompt is for web

MCP.json JSON and this prompt is for web and UA test case. This is for page object model and this is for API test.

Okay.

So that's it about MCPS copilot how we can use AI concepts in ID. Pycharm ID

especially supported. So earlier almost one two months back ID is not supported this actually but nowadays it is started supporting PyCharm ID. So it has

supported other ids like uh intellig IDE, visual studio code ID but recently they have given support to PyCharm IDE also. So you can just play with them and

also. So you can just play with them and you can understand better. Okay. So

that's it for MCP AI in automation.

We'll continue in the next class.