🎯 O time que criou o *Claude* (em parceria com Google Cloud) mostrou na prática como usar prompts avançados, Claude Code, MCP Servers, Agent Development Kit (ADK), Vertex AI Agent Engine e o protocolo Agent-to-Agent (A2A) para construir e deployar aplicações completas em minutos.
✅ Do zero ao produção:
Criação de multi-agentes (ex: Birthday Planner)
Integração com ferramentas via MCP
Deploy serverless com zero DevOps
Orquestração de agentes colaborativos
*Duração original:* ~26 minutos
*Legendado em português por @nett0eth* (muito obrigado pela acessibilidade!)
🔗 Links importantes:
*Vídeo original (inglês) - Anthropic / Google Cloud*
• Building AI agents with Claude in Google C...
*Post original no X (com legenda PT)*
https://x.com/nett0eth/status/2056897...
*Speaker:* Ivan Nardini – Developer Relations Engineer, AI/ML @ Google Cloud
🛠️ O que você vai aprender:
Como o Claude acelera todo o ciclo de desenvolvimento (PM, UI/UX, Engenheiro, Segurança, Marketing)
Stack completa de agentes no Google Cloud (ADK + MCP + Agent Engine + A2A)
Dicas reais de prompting usadas em produção
Deploy de agentes escaláveis com monitoramento e governança
Se você quer ir além do básico de prompts e realmente **aproveitar o poder do Claude**, esse vídeo é ouro! 🔥
👉 Curtiu? Deixa o like, se inscreve e ativa o sininho para mais conteúdos legendados de IA e Crypto!
#Claude #GoogleCloud #VertexAI #AgentesIA #Anthropic #IA #PromptEngineering #ClaudeCode #MCP
✅ Do zero ao produção:
Criação de multi-agentes (ex: Birthday Planner)
Integração com ferramentas via MCP
Deploy serverless com zero DevOps
Orquestração de agentes colaborativos
*Duração original:* ~26 minutos
*Legendado em português por @nett0eth* (muito obrigado pela acessibilidade!)
🔗 Links importantes:
*Vídeo original (inglês) - Anthropic / Google Cloud*
• Building AI agents with Claude in Google C...
*Post original no X (com legenda PT)*
https://x.com/nett0eth/status/2056897...
*Speaker:* Ivan Nardini – Developer Relations Engineer, AI/ML @ Google Cloud
🛠️ O que você vai aprender:
Como o Claude acelera todo o ciclo de desenvolvimento (PM, UI/UX, Engenheiro, Segurança, Marketing)
Stack completa de agentes no Google Cloud (ADK + MCP + Agent Engine + A2A)
Dicas reais de prompting usadas em produção
Deploy de agentes escaláveis com monitoramento e governança
Se você quer ir além do básico de prompts e realmente **aproveitar o poder do Claude**, esse vídeo é ouro! 🔥
👉 Curtiu? Deixa o like, se inscreve e ativa o sininho para mais conteúdos legendados de IA e Crypto!
#Claude #GoogleCloud #VertexAI #AgentesIA #Anthropic #IA #PromptEngineering #ClaudeCode #MCP
Categoria
🤖
TecnologiaTranscrição
00:00Hello, everyone.
00:03Thank you for being here today.
00:05I'm pretty much excited to have this session with you.
00:08I know we are almost at the end of the event,
00:11so thank you for bearing with us.
00:13So just to start, let me introduce myself.
00:16I'm Ivan Nardini.
00:17I'm a developer advocate at Google Cloud
00:19working on building content in partnership with Anthropic.
00:23And in order to start this presentation today,
00:25I just want to ask a very simple question.
00:30So how many of you in the last week
00:32used any AI tool to code or build application?
00:36Okay, the majority.
00:39How many of you used the same AI tool
00:41to build and deploy application on Google Cloud?
00:46Yeah, just a few of them.
00:47So the goal here today is just try to make it better.
00:51So in this demo, what I'm going to show you
00:54is how you can use cloud on Google Cloud
00:58to build and deploy application end-to-end.
01:02And in order to do that,
01:04I'm going to wear five different hats.
01:07And we will start from imagine the use case, right?
01:12Imagine the scenario.
01:13You are an enterprise context.
01:15In an enterprise context,
01:16you probably have a team like the one
01:18that we are visualizing here
01:20that it's engaged to build a new feature
01:24or a new product.
01:25So starting from the left,
01:27you probably have a PM.
01:29Some of you in this room are PMs.
01:31And the PM might have an idea
01:33of how to improve a particular product
01:36or how to implement a feature.
01:38And starting from this idea,
01:41he shared this idea with a UI UX design
01:44which allows him to design the idea,
01:47to visualize the idea.
01:48And after the idea started getting a shape,
01:52then the idea is sent to a software engineer
01:56that essentially started developing
01:57the core logic behind this idea
02:00in order to ship
02:01and make the application accessible
02:03or the feature accessible to everyone.
02:09But before to ship it,
02:10in order to be confident
02:11on what you are releasing,
02:12you probably have to pass a security review.
02:16So that's why you have a role
02:18like the security engineer
02:20that will allow you to be confident
02:22in the release.
02:23And finally,
02:24once the application gets released,
02:27you probably have a data persona
02:31like a growth marketer
02:32or a data analyst
02:33that analyze the data
02:36that are collected through the app
02:37in order to generate insight
02:39and produce feedback to the PM
02:41in a way that the product
02:44or the feature itself gets improved.
02:48So these are kind of, you know,
02:49all the personas
02:50that in a simple way
02:52we can imagine are involved
02:53in the software development lifecycle.
02:56Now, with respect to these personas,
02:59Claw Code,
03:00the Anthropics coding agent,
03:02augment all of them
03:04providing several components
03:06that we are going to use today.
03:08So, as I said,
03:11for the remaining part
03:12of this presentation,
03:13I'm going to put the hat
03:14of all these personas
03:15and I'm going to show you
03:16how you can use
03:17these Cloud Code components
03:19together with Cloud
03:21on Google Cloud
03:22in order to ship,
03:24build and ship
03:25a very simple feedback application
03:27that at the end
03:28of the presentation
03:29we are going to use
03:29to rate my performance here.
03:35But before to start building,
03:37of course,
03:38you need to set up
03:40the Cloud Code
03:42that we are going to use.
03:43And I'm so excited
03:45and so proud
03:46that we work together
03:47with Anthropics
03:48to make this process
03:49of setting up Cloud Code
03:51in order to use
03:52model on Google Cloud
03:54in a very simple way,
03:55in a very straightforward way.
03:57So, you have multiple methods
03:59to use Cloud models
04:01on Google Cloud
04:02in Cloud Code,
04:02but the simplest one,
04:04the faster one,
04:05is using the application
04:07default credential,
04:08which automatically
04:10finds your credential,
04:12for example,
04:13the user one,
04:14and based on the environment
04:16that you are going to use.
04:17And as you can see
04:18in this representation,
04:20recently,
04:20Cloud Code also introduced
04:22this wizard
04:23that will simply
04:24allows you to detect
04:26your project
04:27and your region
04:28where the models
04:29are served
04:29and check
04:32which models
04:32are available
04:33in your project
04:34and let you
04:37to pin them
04:37in order to start
04:38building your application.
04:40At this point,
04:41probably you are familiar
04:43with this
04:44and you are wondering,
04:45okay,
04:45but what is different
04:47to use just Cloud Code
04:48with Cloud models?
04:50Why using Cloud
04:52on GCP,
04:53on Google Cloud?
04:54So, there are many reasons
04:55why you want to do this.
04:58So, first of all,
05:00because you pay
05:01for what you use.
05:02So, the usage
05:04of a Cloud model
05:05from Google Cloud
05:06is per token.
05:07So, you don't receive
05:08a message cap.
05:10And also,
05:11if you are building
05:12an enterprise application
05:14that needs to go
05:16to production,
05:17you can always access
05:19to what is called
05:20provision in throughput,
05:21which essentially
05:22will reserve
05:22some throughput
05:24for you
05:25in order to build
05:26this kind of application.
05:28The other important reason
05:30why you want
05:30to consider
05:32Cloud on Google Cloud
05:33is, as I said,
05:35the setup
05:35is pretty straightforward
05:36using the ADC.
05:38You don't have
05:39an API to rotate
05:40or, you know,
05:42environment variable
05:43to set
05:44in some sense.
05:45So,
05:46it's a happy journey
05:48with respect
05:50to this aspect.
05:51You can access
05:52model
05:52in your project
05:54with your own,
05:55you know,
05:56policies set.
05:58And also,
05:59like,
06:00the data stays
06:00in your project
06:01while you are
06:02interacting
06:02with the Cloud Code.
06:04And model
06:04are served
06:05in multiple regions.
06:07So, you have
06:07global endpoint,
06:09you have regional
06:09endpoint,
06:10depending on,
06:11you know,
06:11the availability
06:12that you need.
06:13And as a Google Cloud,
06:14talking about availability,
06:16we have very great
06:17availability service
06:19standards
06:20that will allow you
06:22to use Cloud
06:24and on one
06:26of the most performing
06:27infrastructure
06:27that you can find
06:29in the market.
06:30So,
06:31these are all
06:32some of the main reasons
06:33why you want
06:34to consider Cloud
06:35on Google Cloud,
06:36especially in an
06:37enterprise contest.
06:38So,
06:38now that you have,
06:39like,
06:40few reasons
06:40of why using Cloud
06:42on Google Cloud,
06:43we are ready
06:44to build.
06:46And so,
06:46as I said,
06:47I will start
06:47wearing the hat
06:48of a PM.
06:50So,
06:50imagine that
06:51you just joined
06:52the company
06:53or maybe you are
06:54already part
06:54of the company.
06:55You have,
06:56you want to improve
06:57our services,
06:57you want to implement
06:58a new features
07:00with respect
07:01to a particular product.
07:02What was happening
07:03in the past
07:04is that you have
07:05the idea,
07:06you go to
07:07a UI UX designer
07:08and you ask him
07:10to prototype
07:10and visualize
07:11the idea.
07:12Now,
07:13with the Cloud
07:14and Cloud Code,
07:16what you can do
07:17is just
07:19drawing a picture
07:20like the one
07:20that you see
07:21here
07:22while you are
07:23drinking a coffee
07:23maybe in San Francisco
07:25and then let
07:27Cloud
07:28doing,
07:29implementing
07:30the idea
07:30for you.
07:31So,
07:32let's see
07:32this in action.
07:37So,
07:38this is
07:38the Cloud Code
07:40UI,
07:41like,
07:41you will probably
07:42familiar with that
07:43and you are
07:43familiar with the
07:44Cloud MD
07:44which is essentially
07:45gives some instruction.
07:47Here,
07:47we just say
07:48that we are a PM,
07:50we want to have,
07:52starting from the picture,
07:54we want to render
07:56a prototype
07:57of the app,
07:59the wireframe
07:59that we are going
08:00to then use
08:01and pass
08:02to the UX designer
08:04and in few minutes
08:05you can see
08:06how Cloud
08:07was capable
08:07of rendering it
08:09and
08:10just starting
08:11from a very
08:12simple picture
08:13or drawing
08:14that you
08:16did
08:17while you were
08:17drinking your coffee.
08:19So,
08:20pretty straightforward.
08:21But imagine
08:22how much time
08:23you save
08:24in doing this
08:24because,
08:25compared to
08:26what you were doing
08:26in the past
08:27with the back and forth
08:28in order to get
08:30this first prototype
08:31of your idea.
08:34Okay.
08:35So,
08:36at this point,
08:37the PM
08:39creates a prototype
08:40and passes
08:41these prototypes
08:42to the UI UX
08:44developer
08:45and at this point
08:47it needs to implement
08:48a more solid
08:49interface
08:50in order to use it
08:51in production.
08:53So,
08:54in this particular
08:55use case,
08:56what we want
08:57to create
08:57is at least
08:58three pages
09:00from the landing
09:01to the tanking
09:03message,
09:04like message page
09:05and a dashboard
09:07that will allow me
09:09to show you
09:09in real time
09:10what can be
09:11the feedback
09:12that I will receive
09:13from the room.
09:14So,
09:14in this case,
09:16there are many ways
09:17you can implement this
09:19but in this case
09:20I want to use
09:21an additional capabilities
09:22of code
09:22which is
09:24the plan mode.
09:25So,
09:25with the plan mode
09:26what we do
09:27we put cloud
09:27in a mode
09:28where
09:29it thinks
09:30before to
09:31like
09:32it thinks
09:32and proposes
09:33what it's going
09:34to do
09:34before to implement
09:35any code
09:37and this is very
09:37important
09:38because it gives me
09:39like a degree
09:40of freedom
09:41of deciding
09:42to change something
09:43according to
09:44my preference
09:45or according
09:45to some standard
09:46that probably
09:48I will get access
09:49to an MCP server
09:50using Figma
09:51for example.
09:53So,
09:54now that we have
09:55in mind
09:55what we are going
09:56to build
09:56let's see this
09:57in action as well.
09:58so we started
09:59from the wireframe
10:01from the PM
10:02similar prompt
10:04I enabled
10:04the plan
10:05the plan mode
10:06and so
10:07as you can see
10:08compared to before
10:09in this case
10:10I'm simulating
10:11the receiving
10:12some instruction
10:13from Figma
10:14using a
10:14design doc
10:15but as you can see
10:17compared to before
10:18it creates a plan
10:19of what it's going
10:20to do
10:20with respect
10:21to all the components
10:22that are defined
10:23in the slide
10:24we look at them
10:25we are happy
10:26we accept
10:27and Cocode
10:28will implement
10:29all of them
10:30and at the end
10:31what we get
10:32is
10:33this
10:34optimized version
10:36so as you can see
10:37we start from here
10:38and
10:39we get this
10:41very
10:42very straightforward
10:43but you can see
10:44how
10:44we are
10:45shifting
10:46from a prototype
10:47to something
10:48that can be
10:49used
10:49in this session
10:51in a very simple way
10:54ok
10:55so this is the part
10:56that probably
10:57every view
10:58or every
10:58of you
10:59in this room
11:00like
11:00you do every day
11:01right
11:03let's
11:03let's wear
11:04the third hat
11:05which is the one
11:05of the
11:06software engineer
11:08and
11:08the software engineer
11:11you receive
11:12these front end
11:13like
11:14all the components
11:15that I was sharing
11:16with you before
11:16and
11:18maybe
11:19he doesn't know
11:19anything
11:20as probably
11:21some of you
11:22in this room
11:23it doesn't know
11:24how to deploy
11:25this application
11:25on Google Cloud
11:27right
11:27so
11:30how you can
11:31how you
11:32can do that
11:33how you can
11:35you know
11:35hand to have
11:37this clear picture
11:38of what are the
11:39components
11:39on Google Cloud
11:41that you need to use
11:42in order to
11:42deploy a very simple
11:44application
11:44like the one
11:45that I show you
11:46today
11:47luckily
11:48it's not a problem
11:49because as a Google Cloud
11:50we invest a lot
11:52of time
11:52to integrate
11:54with these
11:54large
11:55by coding
11:56ecosystem
11:57that now
11:57is
11:57growing
11:59around
12:00models
12:01and
12:02we have
12:03in the last few months
12:05we released
12:06two
12:06important components
12:08so the first one
12:10is
12:10the developer
12:11knowledge API
12:12with
12:13the associated
12:14MCP server
12:15and the second one
12:16is the Google
12:17Cloud skills
12:18so with the
12:19knowledge
12:20with the developer
12:21knowledge API
12:22you get access
12:23to a fresh
12:24documentation
12:25from Google Cloud
12:27that can be
12:27directly consuming
12:28cloud code
12:29through the
12:30MCP server
12:31and it will have
12:33cloud code
12:33to figure it out
12:34what is the best
12:36architecture
12:37what is the best
12:38implementation
12:39to deploy a certain
12:40application
12:41on Google Cloud
12:42this is very
12:43important
12:43because again
12:44what we are
12:45saying here
12:45is that
12:46you don't need
12:46to know
12:47like how to
12:48deploy an
12:48application
12:49on Google Cloud
12:50you can just
12:50leverage
12:51cloud code
12:52and this
12:52MCP server
12:53that we expose
12:54now on
12:55Google Cloud
12:56side
12:56to build
12:57applications
12:58like this one
12:58so in this
13:00case
13:00probably
13:01what we are
13:02going to do
13:03is that
13:03we are going
13:03to deploy
13:04the feedback
13:05API
13:05on a serverless
13:06function
13:07like a
13:07cloud run
13:08we will
13:08connect
13:09with a
13:10web
13:10oriented
13:12DB
13:12like a
13:13firestore
13:13to collect
13:15the raw
13:15responses
13:16that we
13:16will give
13:17through the
13:18feedback app
13:19and then
13:20because we
13:20want to have
13:21that analytics
13:22part
13:22we will
13:23build
13:24implementation
13:25in a way
13:26that we
13:27can store
13:27those raw
13:28response
13:29in an
13:29analytical
13:30data warehouse
13:30like
13:31BigQuery
13:32and
13:33in BigQuery
13:34we will
13:35post-process
13:35and we
13:36consume
13:36this
13:37information
13:37in a
13:38dashboard
13:41like
13:42the one
13:42that
13:43you
13:43can
13:43find
13:43in
13:44Looker
13:44but
13:45again
13:46you
13:47can
13:48build
13:48this
13:48using
13:49a
13:51code
13:51in
13:51combination
13:52with
13:52the
13:52MCP server
13:53without
13:54you having
13:55prior knowledge
13:56of how
13:57to do
13:57that
13:58this
13:59like
14:00the MCP
14:00server
14:00and
14:01the
14:01developer
14:02knowledge
14:02API
14:03that I
14:03just
14:03mentioned
14:03it
14:04got
14:04paired
14:04with
14:04also
14:05the
14:05skill
14:05part
14:06and
14:07the
14:07skill
14:07part
14:07is
14:08if
14:08with
14:08the
14:09MCP
14:09server
14:10you
14:10will
14:10be
14:10able
14:11to
14:11design
14:11the
14:11architecture
14:12with
14:12the
14:13skills
14:13you
14:13will
14:14be
14:14able
14:14to
14:14cover
14:14the
14:15single
14:15blocks
14:15of
14:16this
14:16architecture
14:16for
14:16example
14:17we
14:17released
14:17a
14:18simple
14:19skills
14:19that
14:20enable
14:21cloud
14:22to deploy
14:23on
14:23cloud
14:24run
14:24to
14:24deploy
14:25an
14:25API
14:25on
14:25cloud
14:26run
14:26or
14:26to
14:27connect
14:28like
14:29cloud
14:30run
14:30with
14:30fire
14:30store
14:31so
14:31it's
14:32more
14:32about
14:33the
14:33implementation
14:34itself
14:35rather
14:35than
14:36you
14:36know
14:36giving
14:36the
14:37overall
14:37picture
14:40so
14:41once
14:42you
14:42get
14:42cloud
14:43code
14:43enabled
14:43with
14:44the
14:44MCP
14:45server
14:46of
14:47the
14:47documentation
14:47and
14:48the
14:48skills
14:48that
14:48I
14:49just
14:49mentioned
14:49you
14:50can
14:50just
14:51directly
14:51implement
14:52the
14:53architecture
14:54that I
14:54was
14:54showing
14:55you
14:55before
14:55and
14:56you
14:56can
14:56do
14:56this
14:57in
14:57parallel
14:57so
14:58another
14:58component
14:59that
14:59you
14:59can
14:59use
15:00in
15:00cloud
15:00code
15:00is
15:01a
15:01sub
15:01agents
15:01and
15:02as
15:02you
15:02can
15:02see
15:03here
15:03we
15:03are
15:03going
15:04to
15:04spin
15:04up
15:04three
15:04different
15:05sub
15:05agents
15:05one
15:06for
15:06the
15:06API
15:07one
15:07for
15:07the
15:08ingestion
15:09pipeline
15:09and
15:09the
15:09other
15:10one
15:10for
15:10the
15:10dashboard
15:11and
15:11you
15:11can
15:12parallelize
15:12the
15:13implementation
15:13of each
15:14of
15:14these
15:14components
15:14just
15:15like
15:16you
15:16run
15:17a
15:17team
15:18sprint
15:18in
15:19your
15:20normal
15:20like
15:21usual
15:21development
15:22lifecycle
15:24so
15:25let's
15:25see
15:25this
15:25in
15:25action
15:26as
15:26well
15:28so
15:29here
15:29we
15:30are
15:30again
15:30first
15:31of
15:31all
15:31I
15:31just
15:32want
15:32to
15:32show
15:32you
15:32that
15:33we
15:33have
15:33enabled
15:33the
15:34mcp
15:34server
15:35of
15:35the
15:35documentation
15:35and
15:36we
15:36have
15:37the
15:37skills
15:37some
15:37of
15:38the
15:38skills
15:38that
15:38we
15:39pre-built
15:39one
15:41time
15:41more
15:41I
15:42provide
15:42a
15:42very
15:42simple
15:43prompt
15:43the
15:44first
15:44step
15:44is
15:45designing
15:45the
15:46cloud
15:46native
15:46backend
15:47so
15:48it
15:49will
15:49start
15:49like
15:50it
15:51will
15:51provide
15:51me
15:53draft
15:53of
15:53the
15:54architecture
15:54in
15:54this
15:55case
15:55I
15:55could
15:55use
15:56again
15:56the
15:56plan
15:56mode
15:57but
15:57for
15:57simplicity
15:58I
15:58didn't
15:58and
15:59then
15:59use
15:59the
16:00skill
16:00one
16:01of
16:01the
16:01skill
16:01that
16:01I
16:01provide
16:02in
16:02order
16:02to
16:02implement
16:03the
16:03API
16:04let's
16:04say
16:05that
16:05we
16:05are
16:05happy
16:06with
16:06the
16:06API
16:06spec
16:07then
16:08we
16:08have
16:09the
16:09architecture
16:10we
16:10have
16:10the
16:10API
16:11spec
16:11the
16:12next
16:12step
16:12is
16:13running
16:13multiple
16:14agents
16:15in
16:15parallel
16:16in
16:16order
16:16to
16:16implement
16:18the
16:18three
16:19components
16:19of
16:19the
16:20app
16:20that
16:20I
16:21was
16:21showing
16:21you
16:21so
16:23it's
16:23pretty
16:24quick
16:24as
16:25you
16:25know
16:25this
16:25code
16:26like
16:27it
16:27also
16:27managed
16:28the
16:28testing
16:28part
16:29after
16:30it
16:30finished
16:30the
16:30implementation
16:31and
16:31at
16:32the
16:32end
16:32you
16:33will
16:33get
16:33your
16:34app
16:35which
16:35is
16:35now
16:36ready
16:37to
16:37deploy
16:38on
16:38Google
16:39Cloud
16:43okay
16:44so
16:45at
16:45this
16:45point
16:47we
16:47have
16:47the
16:48code
16:49of
16:49the
16:49app
16:50that
16:50is
16:50ready
16:51to
16:51be
16:51deployed
16:51but
16:52because
16:52we
16:53are
16:53deploying
16:53on
16:54cloud
16:54and
16:54we
16:54want
16:54to
16:55open
16:55this
16:55up
16:55to
16:56a
16:56larger
16:56audience
16:57we
16:57want
16:57to
16:57deploy
16:58it
16:58confidently
16:59like
17:00so
17:02this
17:03is
17:03when
17:04you
17:04want
17:04to
17:04consider
17:05to
17:05run
17:06a
17:06security
17:06review
17:07now
17:07depending
17:08on
17:08your
17:09company
17:09you
17:10can
17:11have
17:11different
17:12security
17:13requirements
17:13so
17:14for
17:14example
17:14you
17:15may
17:15want
17:15to
17:15check
17:15if
17:16your
17:16application
17:17is
17:18solid
17:18with
17:18respect
17:19to
17:19the
17:19most
17:20common
17:22OWASP
17:23issue
17:24or
17:25because
17:26you
17:26are
17:26deploying
17:26this
17:27application
17:27on
17:27cloud
17:28and
17:28one
17:29thing
17:29that
17:30you
17:30need
17:30to
17:30consider
17:30is
17:31probably
17:31you
17:32will
17:32use
17:32what
17:33is
17:33called
17:33a
17:33service
17:34account
17:34and
17:35you
17:35want
17:35to
17:35limit
17:36the
17:37service
17:38account
17:38when
17:39it
17:39calls
17:39a
17:40particular
17:40API
17:40like
17:41the
17:41one
17:41related
17:42to
17:42reading
17:43and
17:43writing
17:43a
17:43DB
17:43with
17:44respect
17:44to
17:44certain
17:45role
17:45so
17:46you
17:46are
17:46you
17:46you
17:47you
17:47sure
17:47that
17:48you
17:49are
17:49limiting
17:49what
17:50it
17:50can
17:50happen
17:50when
17:51when
17:51the
17:52application
17:52runs
17:53some
17:53operation
17:53on
17:54the
17:54cloud
17:54itself
17:54so
17:55again
17:56these
17:57are
17:57just
17:57a
17:57couple
17:57of
17:57examples
17:58of
17:59what
18:00you
18:00want
18:00to
18:00consider
18:01in
18:01a
18:01phase
18:02like
18:03this
18:03one
18:03and
18:04of
18:04course
18:04this
18:05representation
18:05is a
18:06strong
18:07simplification
18:07of what
18:08can happen
18:09in the
18:09real life
18:11it's just
18:11one of the
18:12possible
18:12scenario
18:13that you
18:14can have
18:14and you
18:15can see
18:15why
18:16it's also
18:16simplification
18:17because
18:17we are
18:18letting
18:18the
18:19security
18:19engineer
18:20not only
18:20to
18:21approve
18:21if the
18:22app
18:22is
18:22secure
18:23enough
18:23to be
18:23deployed
18:24but
18:24it
18:24will
18:24also
18:25deploy
18:25the
18:25app
18:25in
18:26this
18:26case
18:26but
18:27again
18:28this
18:28is
18:29just
18:29a
18:29demo
18:29so
18:29we
18:30have
18:30this
18:30degree
18:31of
18:31freedom
18:31with
18:32that
18:32being
18:32said
18:32let's
18:33run
18:33the
18:34final
18:34demo
18:34and
18:35let's
18:35get
18:36the
18:36app
18:36running
18:37so
18:38in
18:39this
18:39case
18:39as
18:40you
18:40can
18:40see
18:40I
18:40use
18:41a
18:41pre-built
18:41security
18:42review
18:42that
18:43you
18:43can
18:43find
18:43in
18:43Cocode
18:45very
18:45simple
18:47prompt
18:48also
18:48in this
18:49case
18:49and
18:50what
18:51is
18:51happening
18:51is
18:51Cocode
18:52run
18:52the
18:52first
18:53test
18:53essentially
18:54double
18:55check
18:55that
18:57everything
18:57is
18:58aligned
18:58it
18:58found
18:59like
18:59a
18:59possible
19:00issue
19:00and
19:01it
19:02automatically
19:02fix
19:02it
19:03so
19:03in
19:04a
19:04way
19:04that
19:04the
19:04app
19:04now
19:05is
19:05secure
19:05and
19:06once
19:06it
19:06is
19:06secure
19:07it
19:07deployed
19:07the
19:08back
19:09end
19:09API
19:10and
19:10it
19:10deployed
19:11the
19:11API
19:11itself
19:12so
19:13what
19:13we
19:13get
19:14at
19:14the
19:14end
19:14of
19:15this
19:15it's
19:15an
19:16endpoint
19:16with
19:16our
19:16app
19:17the
19:18app
19:18is
19:18live
19:18so
19:19I
19:19will
19:21quickly
19:21unlock
19:23my
19:23laptop
19:23and
19:24I
19:24will
19:25ask
19:27the
19:28backstage
19:29to
19:29share
19:30on my
19:30laptop
19:32so
19:33the
19:33app
19:34as you
19:34can
19:34see
19:35is
19:35up
19:35and
19:35running
19:36on
19:36GCP
19:36for people
19:37that
19:38doesn't
19:38know
19:38GCP
19:38this
19:39is
19:39cloud
19:39run
19:40it's a
19:40serverless
19:41service
19:42that you
19:43can use
19:43in order
19:43to deploy
19:44app
19:44and
19:45the
19:46app
19:46at the
19:47end
19:47looks
19:48like
19:48this
19:48one
19:48so
19:49if
19:50you
19:50remember
19:50I
19:51show
19:51you
19:51the
19:51original
19:52I
19:53show
19:53you
19:54the
19:54feedback
19:55frame
19:56at the
19:56beginning
19:57so
19:58what
19:58we
19:58can
19:58do
19:58now
19:59live
19:59I
20:00can
20:00just
20:00give
20:00me
20:01a
20:01score
20:01what
20:02do
20:02you
20:02think
20:02the
20:02session
20:02is
20:03going
20:03so
20:03far
20:03give
20:04me
20:05five
20:06oh
20:07okay
20:07thank
20:07you
20:08man
20:08I
20:08really
20:08appreciate
20:09that
20:10okay
20:11cool
20:11session
20:12let's
20:12be
20:13simple
20:14I
20:14submit
20:15and
20:16then
20:16in
20:17real
20:17time
20:17it
20:18updates
20:18the
20:18number
20:18of
20:19response
20:19the
20:20score
20:20and
20:21you
20:21know
20:21the
20:22visualization
20:22and
20:23also
20:23just
20:23for
20:23fun
20:24I
20:24build
20:25a
20:25feedback
20:25analyzer
20:26so
20:27once
20:27I
20:27click
20:27this
20:28it
20:28will
20:29run
20:29it
20:29will
20:30call
20:30cloud
20:32on
20:32Google
20:33cloud
20:33and
20:34based
20:34on
20:34the
20:35feedback
20:35and
20:35the
20:35comments
20:36that
20:36you
20:36send
20:36it
20:37will
20:37generate
20:37this
20:37summary
20:39so
20:40pretty
20:41straightforward
20:45I
20:46will
20:46ask
20:46to
20:46go
20:46back
20:46on
20:47the
20:47presentation
20:47thank
20:48you
20:50okay
20:50so
20:51at
20:51this
20:51point
20:52we
20:52have
20:52the
20:52app
20:53we
20:53collect
20:54good
20:55feedback
20:55thank
20:55you
20:56again
20:56man
20:57but
20:58the
20:58development
21:00lifecycle
21:00is still
21:01there
21:01we
21:02have
21:02the
21:02last
21:02step
21:03to
21:03cover
21:03which
21:04is
21:04essentially
21:04people
21:05there
21:05start
21:06using
21:07our
21:07app
21:07if
21:08you
21:08saw
21:08one
21:08of
21:09the
21:09KPI
21:09that
21:09I
21:10had
21:10on
21:10the
21:10dashboard
21:10was
21:11the
21:11time
21:11the
21:12response
21:12time
21:12so
21:13how
21:13long
21:13was
21:14taking
21:15you
21:15to
21:15just
21:16in
21:17providing
21:18feedback
21:18so
21:19these
21:19are
21:20the
21:20kind
21:20of
21:20information
21:20that
21:21you
21:21can
21:21use
21:21in
21:22order
21:23to
21:24through
21:25this
21:25data
21:25you
21:26can
21:26collect
21:27this
21:27data
21:27analyze
21:28them
21:28and
21:28generate
21:29insights
21:30in
21:30a way
21:30that
21:31they
21:31can
21:31be
21:31used
21:32in
21:32order
21:32to
21:32improve
21:32the
21:33application
21:34now
21:36running
21:37if you're
21:38new on
21:38Google Cloud
21:39there are
21:39several services
21:40that you can
21:41use
21:41in order
21:41to
21:42analyze
21:42this
21:42data
21:43that
21:43comes
21:43from
21:43the
21:44app
21:44so
21:44a couple
21:46of
21:46them
21:46one
21:47as I said
21:47at the
21:47beginning
21:48is
21:48an
21:48analytical
21:49data
21:49warehouse
21:49which
21:50is
21:50BigQuery
21:50and
21:51for
21:52the
21:52reporting
21:52part
21:53you
21:53have
21:54a
21:54tool
21:54like
21:54Looker
21:55but
21:55again
21:56as we
21:56said
21:57before
21:57you
21:58don't
21:58need
21:58to
21:58know
21:59how
21:59to
21:59use
21:59BigQuery
22:00in
22:00order
22:00to
22:01analyze
22:01those
22:02data
22:02as well
22:02as
22:03how
22:03to
22:03build
22:04a
22:04dashboard
22:04in
22:04Looker
22:05because
22:05we
22:06also
22:07provide
22:07MCP
22:08server
22:08for
22:09doing
22:09this
22:09now
22:10for
22:10the
22:10sick
22:10of
22:10time
22:11I'm
22:11not
22:11going
22:11to
22:11demo
22:11how
22:12to
22:12use
22:13an
22:13MCP
22:13server
22:14to
22:14query
22:14BigQuery
22:15or
22:15building
22:15a
22:16dashboard
22:16but
22:17I
22:17want
22:18to
22:18quickly
22:18share
22:18with
22:19you
22:20where
22:21you
22:21can
22:21find
22:22this
22:22information
22:22in order
22:23to
22:23do
22:23that
22:24after
22:24this
22:24session
22:25because
22:25we
22:25are
22:26going
22:26to
22:26release
22:26the
22:26code
22:27so
22:27I
22:27will
22:28ask
22:28to
22:29shift
22:30back
22:30on
22:30my
22:30laptop
22:31thank
22:31you
22:32so
22:36loading
22:38time
22:38so
22:39with
22:39respect
22:40to
22:40all
22:40the
22:40MCP
22:41server
22:41that
22:42are
22:42available
22:42on
22:43Google
22:43Cloud
22:43we
22:44recently
22:45announced
22:45the
22:46agent
22:46platform
22:47so
22:47in
22:48the
22:48agent
22:48platform
22:49you
22:49have
22:50services
22:51that
22:51is
22:52represented
22:52by
22:53the
22:53agent
22:53registry
22:53and
22:54in
22:54the
22:54agent
22:54registry
22:55you
22:55have
22:55the
22:56list
22:56of
22:57all
22:57the
22:57MCP
22:57server
22:58that
22:58we
22:58natively
22:59support
22:59on
22:59Google
23:00Cloud
23:00so
23:00for
23:00example
23:01we
23:01have
23:01the
23:02developer
23:02knowledge
23:02service
23:03that
23:04I
23:04just
23:04show
23:04you
23:04and
23:05we
23:06also
23:08the
23:08bQuery
23:09MCP
23:09server
23:10that
23:10you
23:10can
23:11use
23:12in
23:12order
23:13to
23:14query
23:15the
23:15data
23:15that
23:16we
23:16just
23:16collected
23:16from
23:17the
23:17app
23:18it's
23:19very
23:19like
23:19this
23:20registry
23:20is
23:20relevant
23:21because
23:21it
23:21tells
23:21you
23:22how
23:22to
23:22set
23:23the
23:24MCP
23:25server
23:25on
23:26your
23:26side
23:26as
23:27well
23:27as
23:27it
23:27gives
23:28you
23:28some
23:28observability
23:29feature
23:29and
23:30the
23:30description
23:31of
23:31all
23:31the
23:31tools
23:32that
23:32you
23:32will
23:32find
23:33for
23:33the
23:34MCP
23:34server
23:34so
23:34you
23:35know
23:35how
23:35code
23:36will
23:36be
23:38this
23:38server
23:40to
23:40query
23:41your
23:41data
23:42with
23:42respect
23:43to
23:43the
23:45looker
23:46part
23:46also
23:47like
23:47we
23:49released
23:50the
23:50MCP
23:50toolbox
23:51of
23:51DB
23:52this
23:52is
23:52a
23:52open
23:53source
23:53like
23:54a
23:54model
23:55context
23:55protocol
23:55server
23:56which
23:57include
23:57an
23:58integration
23:58with
23:58looker
23:59and
23:59it's
24:00very
24:00well
24:02we
24:02have
24:02a
24:03very
24:03well
24:04documented
24:04quick
24:04start
24:05on
24:05how
24:06to
24:06set
24:06up
24:06with
24:07the
24:07clock
24:07code
24:08and
24:08start
24:08using
24:09it
24:09in
24:09order
24:09to
24:10consume
24:10that
24:10data
24:10from
24:11BigQuery
24:11and
24:11build
24:12your
24:12dashboard
24:12so
24:13I
24:13leave
24:14you
24:14this
24:14as
24:14an
24:15exercise
24:15like
24:16we
24:16are
24:17going
24:17to
24:17release
24:17the
24:17code
24:18so
24:18you
24:18can
24:18go
24:18home
24:19and
24:19integrate
24:19these
24:20two
24:20parts
24:20is
24:20pretty
24:21straightforward
24:21but
24:22the
24:23dashboard
24:23that
24:23you
24:23can
24:24create
24:24they
24:24are
24:25pretty
24:25powerful
24:26and
24:26you
24:27will
24:27see
24:27how
24:28nice
24:28they
24:28can
24:29be
24:30okay
24:30back
24:31to
24:31the
24:32presentation
24:33I
24:34think
24:34I'm
24:34just
24:35in
24:35time
24:35so
24:35time
24:36to
24:36wrap
24:37up
24:37what
24:38I
24:38try
24:38to
24:38explain
24:39you
24:39today
24:40is
24:41essentially
24:41two
24:42things
24:42so
24:43first
24:43of
24:43all
24:44I
24:44was
24:44trying
24:44to
24:44and
24:45I
24:45hope
24:45I
24:45did
24:46good
24:46enough
24:47I
24:47was
24:47trying
24:48to
24:48show
24:48you
24:48how
24:49like
24:49all
24:50the
24:50components
24:51of
24:51a
24:51clock
24:51code
24:52including
24:52skills
24:53MCP
24:54server
24:54sub
24:55agents
24:56they
24:56can
24:56really
24:56speed
24:57up
24:57the
24:58process
24:58of
24:58software
24:59development
24:59as
25:00well
25:00as
25:01how
25:02you
25:02can
25:02use
25:03clock
25:03code
25:03with
25:04cloud
25:05models
25:05on
25:06GCP
25:06in a
25:07very
25:07seamless
25:07way
25:08like
25:08if
25:08you
25:08saw
25:08we
25:09run
25:09several
25:10sessions
25:10across
25:11like
25:11multiple
25:12personas
25:13and
25:14like
25:15it
25:15was
25:15the
25:16experience
25:16was
25:16just
25:17straightforward
25:18it
25:18was
25:19just
25:19incredible
25:20so
25:21this
25:22is
25:22what
25:23you
25:23can
25:23get
25:23if
25:24you
25:24combine
25:24you
25:25know
25:25clock
25:25code
25:26with
25:26cloud
25:27models
25:27on
25:28GCP
25:28as
25:29I
25:29said
25:31the
25:31code
25:32is
25:33going
25:33to
25:33be
25:34available
25:34right
25:34after
25:35the
25:35session
25:35we
25:35have
25:35a
25:36great
25:36quick
25:36start
25:37and
25:37we
25:37have
25:38a
25:38very
25:38well
25:39maintained
25:40documentation
25:40both
25:41on
25:41Google
25:41call
25:42side
25:42and
25:42Antropic
25:42side
25:43so
25:43I
25:44highly
25:44recommend
25:44to
25:45just
25:45go
25:45there
25:46and
25:46check
25:46out
25:46and
25:47then
25:47I
25:48hope
25:48I
25:49cover
25:49everything
25:49but
25:50if
25:50you
25:50still
25:51have
25:51questions
25:51or
25:52you
25:52want
25:52to
25:52provide
25:52additional
25:53feedback
25:53just
25:54feel
25:54free
25:54to
25:54reach
25:55out
25:55these
25:55are
25:56my
25:56social
25:56media
25:58point
25:59so
26:00with
26:00that
26:00being
26:00said
26:01thank
26:01you
26:02so
26:02much
26:02and
26:02it
26:02was
26:03a
26:03pleasure
26:03being
26:03here
26:04today