
Google Agentspace JUST WON the AI Game...
What this covers
Wes Roth walks through Google's April 2025 announcements at Google Cloud Next 2025, mixing his own commentary with clips from a Google demo video. He starts with a comparison to the original search business: Google made money by sending users off Google.com to sites whose owners paid for placement, unlike rivals that tried to keep people on their own pages. He then treats Agent Space and its open A2A protocol as a possible repeat of that gatekeeping role, with agents in place of links.
The demo material covers Agent Space as a product. It has an agent gallery holding Google's agents, company-built agents, partner agents running on models such as Llama and Claude, and personal agents users create by conversation without code. Those agents can be extended by asking, for example adding an audio summary or email delivery. Other items shown include an enterprise deep research agent that plans, queries Google Search and internal data, and revises its plan as it goes, and a financial time-series forecasting model for cash flow. For banks and similar regulated firms, the pitch is employee-level access controls and operation inside a customer's own VPC.
The longest stretch is on A2A mechanics, illustrated by a multi-week hiring task. A client agent consults an agent registry, uses capability discovery to find agents that advertise skills in simple JSON, and coordinates sourcing, interviews, and background checks. The most instructive moment comes when a US-only background-check agent fails on an international candidate. The client agent records the failure, tells the user, and asks permission to find a different agent. The discussion also covers selection by reputation and past experience, A2A's design principles (secure by default, long-running tasks, modality-agnostic), and the ADK (Agent Development Kit)) for building multi-agent systems.
Google's Agent Space and A2A protocol represent a fundamental shift in how users interact with digital services—from searching for information to delegating tasks to AI agents, potentially recreating Google's dominance in a post-search paradigm where agents become the primary interface to the web.
- Agent Space creates an ecosystem where users interact through personal agents rather than search bars, mirroring Google's original web gatekeeping model
- The A2A protocol enables seamless interoperability between agents from different vendors, creating network effects that could entrench a dominant platform
- Monetization opportunity emerges when service providers pay to have their agents prioritized, paralleling Google's original ad-based business model
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Google Search's success model was based on redirecting users away from Google.com to relevant websites advertised by service providers who paid for placement; this model differs fundamentally from other companies that tried to keep users on their own websites.
“It was very different from how other people build their businesses online because instead of trying to keep them in their sort of website on google.com, it was trying to get them off of google.com, right? So you type in background checks for employment. For most of Google's history, they wanted you off of Google and on to, for example, this ad or this ad or this ad.”
Google has announced Agent Space, a platform that creates an ecosystem where businesses and developers can create agents that seamlessly work together, coupled with an open A2A (agent-to-agent) protocol for developing interoperable agents.
“Google just announced tons of stuff. The thing that really caught my attention is agent space. This thing might be huge and it could be multiple things. Number one, it's an ecosystem where all sorts of businesses and developers are able to create agents and have them seamlessly work together. The other thing that they kind of announced along with this is an open protocol for developing agents. A2A, agent to agent protocol.”
The A2A (Agent-to-Agent) protocol is an open protocol that allows different AI agents built by different vendors or frameworks to communicate with one another, securely exchange information, and coordinate actions to complete tasks.
“This A2A protocol is going to allow these agents to work together, securely exchange information, coordinate actions, etc.”
The A2A protocol enables agents to use capability discovery to advertise their specific capabilities (such as US-based background checks) in a standardized format, allowing client agents to find and select the appropriate remote agent for specific tasks.
“The agents are able to ask for clarifications information and do various sub actions on behalf of the clients or users... The agent replies to you saying here are the three candidates who have passed the interview and are pending background checks... based on my agent registry... I have access to three background check agents using symbol background agent for all candidates given the best experience in the past interactions.”
Agent Space includes an enterprise deep research agent that plans its research approach, pulls real-time information from Google Search, searches internal enterprise data, and adjusts its plan in real-time by adding additional questions based on what it discovers, ultimately providing detailed analysis with source links and executive summaries.
“This activates Google's enterprise deep research agent, which starts by telling me what it plans to research and in what order... This agent is pulling in real-time information from Google Search to build its report, but even cooler, it's also searching my internal enterprise data and adjusting this plan in real time, adding additional questions based on what it's going to find along the way.”
Agent Space's agent gallery allows users to access approved agents created by Google, agents created by their company, agents created by third-party partners using models like Llama and Claude, and personal agents that users create themselves through conversation without coding.
“The agent gallery lets me see my company's approved selection of purpose-built agents, including ones powered by third party models like Llama and Claude. You see, we've got some Google made agents. We have agents that my bank has made available to me, either ones we've created or ones built by partners. And then the best part, my own personal agents, which I can build directly inside agent space with this button over here, or even easier, just from having a little conversation.”
For companies with strict regulatory requirements like banks, Agent Space provides stringent access controls at the employee level and can operate within a customer's own VPC (Virtual Private Cloud), ensuring that customer data remains private while meeting regulatory requirements.
“For companies with strict regulatory needs like a bank, agent, agentspace provides stringent access controls at the employee level and can operate within your own VPC, ensuring that your data stays yours while meeting all of your requirements.”
An agent in the A2A ecosystem can select between multiple available agents providing the same service based on past interaction experience and agent reputation, choosing the agent with the 'best experience in past interactions'.
“I have access to three background check agents using symbol background agent for all candidates given the best experience in the past interactions.”
The A2A protocol enables agents to coordinate complex workflows involving multiple specialized agents, each with narrow, well-defined capabilities, sequencing their work and managing dependencies (e.g., candidate sourcing → interviews → background checks with different agents for US vs international candidates).
“This is fascinating. So it's saying based on my agent registry... I have access to three background check agents using symbol background agent for all candidates given the best experience in the past interactions. This could be huge. I mean imagine this marketplace growing and continuously adding more agents. Each one having their own sort of reputation based on you know users experience with it.”
The A2A protocol design principles emphasize enabling agent collaboration in their natural, unstructured modalities even when agents don't share memory, tools, or context, requiring the protocol to handle heterogeneous environments and communications patterns.
“Here are some of the design principles embrace agent capabilities. So it focuses on enabling agents to collaborate in their natural unstructured modalities even when they don't share memory tools and context.”
The A2A protocol in demonstrated hiring scenario supports multi-week task execution with human involvement, where an agent coordinates candidate sourcing, conducts interviews over multiple days, initiates background checks, tracks completion status, and provides user updates on task progress.
“So next the client and the agent kind of talk back and forth. The agents are able to ask for clarifications information and do various sub actions on behalf of the clients or users... Here, it looks like it found three people that match the job description. It creates a reach out plan and then two weeks later once interviews were complete... the agent replies to you saying here are the three candidates who have passed the interview and are pending background checks.”
Users can create custom agents in Agent Space through conversational interaction without writing code, and the system automatically generates an agent plan based on the conversation and can be refined or extended (e.g., adding audio summary generation and email delivery features).
“And just like that, I have built my own custom agent to use whenever I want without writing a single line of code... Agent Space automatically generates an agent plan based on our conversation so far. And this is good, but I think I want more. I'm gonna have it generate an audio summary and send it to my inbox so I can listen to it on my morning commute.”
Agent Space includes a time series forecasting model specifically trained for financial scenarios, enabling agents to forecast cash flow across future periods (e.g., three quarters) and provide recommended steps based on those forecasts.
“I'm going to have our bank's cash flow agent do some forecasting across the next three quarters for me. This agent uses Google's new time series forecasting model, which is specifically trained for scenarios just like this. And again, I'm going to get a super clear, very, very clear summary with at the bottom some great recommended steps.”
An agent sourcing candidates for a hiring task can identify that certain agents (e.g., US-based background check agents) have scope limitations and automatically discover alternative agents with international capabilities to complete tasks outside the scope of the first agent.
“So it was able to successfully initiate background check for one of the candidates but the other two it failed to do so because they're outside of the US which is outside of the scope for that particular agent. So it goes through and tries to find an international background checking agent and then it was able to initiate the background checks of the remaining two candidates with this particular agent.”
Agent Space contains a repository of pre-built agents including a hiring agent, portfolio analysis agent, cash flow forecasting agent, research agent, and email drafting agent, demonstrating a framework that encompasses both workflow-specific agents and general-purpose agents.
“I'm going to have our bank's cash flow agent do some forecasting... Powered by Gemini 2.5 and Google Search technology, Agent Space is the only hyperscaler platform on the market... Draft me an email to Acme General Contracting CEO”
Agent Space demonstrates an agent-to-agent communication scenario where a client agent requests a hiring task, discovers available agents in the agent registry through 'capability discovery', establishes communication with agents matching the required capabilities, coordinates information exchanges, and delegates work to those agents.
“For example, here we're asking this agent, I am hiring for a staff software engineer position. Attached is the job description. Can you help me fill this role? The agent goes on to uh source candidates and looks through the various available agents for one that's able to do candidate sourcing or as they call it capability discovery. So agents can advertise the their capabilities to clients and then clients can choose to use those specific agents to complete their tasks.”
The A2A protocol demonstrates that when a remote agent lacks the capability to complete a task (such as international background checks), the client agent can discover and invoke a different remote agent with the required capability, automatically switching to handle the new task.
“So it was able to successfully initiate background check for one of the candidates but the other two it failed to do so because they're outside of the US which is outside of the scope for that particular agent. So it goes through and tries to find an international background checking agent and then it was able to initiate the background checks of the remaining two candidates with this particular agent.”
The A2A protocol design follows principles including: embracing agent capabilities by enabling agents to collaborate in unstructured modalities even without shared memory, tools, or context; building on existing standards for easy integration; being secure by default with enterprise-grade authentication; supporting long-running tasks from quick operations to days-long processes; and being modality-agnostic (text, audio, video, streaming).
“So here are some of the design principles embrace agent capabilities. So it focuses on enabling agents to collaborate in their natural unstructured modalities even when they don't share memory tools and context. So it's going to be interesting to see how they sort of pull this off. So it's built on a lot of the existing standards... secure by default with enterprisegrade authentication and authorization support for long running tasks... And it's modality agnostic so it's not just text it can do whatever you want audio video streaming etc.”
Agent Space enables users to refine and extend agent behavior through conversation, for example requesting an agent to add new output modalities (e.g., audio summaries) or change delivery methods (e.g., email delivery) without requiring the agent to be re-trained or re-coded.
“Agent Space automatically generates an agent plan based on our conversation so far. And this is good, but I think I want more. I'm gonna have it generate an audio summary and send it to my inbox so I can listen to it on my morning commute.”
The Agent Development Kit (ADK) is an open-source framework designed to simplify development and deployment of multi-agent systems, providing a quick start guide for installation and integration.
“they're sharing the agent development kit, ADK. It's an open-source flexible framework designed to simplify the development and deployment of multi- aent systems. Here's the quick start for installing that, the ADK, the agent development kit.”
Agents in the A2A registry advertise their capabilities using simple JSON format, enabling capability discovery where agents can list specific capabilities (e.g., 'US-based background checks') and clients can select agents based on matching these declared capabilities.
“So it looks like these agents they can advertise their capability. It's using a simple JSON format. So you kind of list the capabilities like we saw for example US-based background checks right you can say that's one of the capabilities that your agent can do.”
The A2A protocol architecture enables task-oriented communication where the output of a completed task is called an 'artifact', and this architecture is foundational to the Agent Development Kit (ADK) which allows developers to build production-ready multi-agent applications with greater flexibility and control.
“It seems like they're calling the output of a task uh an artifact. And of course, this kind of goes handinhand with the ADK, the agent development kit that allows you to build a multi- aent applications. They introduced it at the current conference, the Google Cloud Next 2025, and it empowers developers to build production ready agentic applications with greater flexibility and precise control.”
Agent Space demonstrates capability discovery by showing an agent querying an agent registry to identify which agents possess specific capabilities (e.g., candidate sourcing) needed to fulfill a user request.
“The agent goes on to uh source candidates and looks through the various available agents for one that's able to do candidate sourcing or as they call it capability discovery.”
When a delegated agent fails to complete a task within its scope of capabilities (e.g., international background checks when it only handles US candidates), the client agent records this failure, communicates it to the user, and requests permission to locate alternative agents capable of completing the failed subtask.
“We tell it to please initiate the background check for all candidates and it goes on to do it... based on my agent registry. So this is the your agent talking to you about the other agents that it has found in the agent space.”
In the A2A protocol architecture, the client agent represents the user and is responsible for secure collaboration, task and state management, user experience negotiation, and capability discovery to select appropriate remote agents.
“You have your client agent representing you. You have the remote agent and your client agent does the secure collaboration task and state management, user experience negotiation, capability discovery.”
Agent Space's deep research agent can access real-time information from Google Search, integrate internal enterprise data, and dynamically adjust its research plan based on findings, allowing it to pull information from both external and internal sources simultaneously.
“This agent is pulling in real-time information from Google Search to build its report, but even cooler, it's also searching my internal enterprise data and adjusting this plan in real time, adding additional questions based on what it's going to find along the way.”
Agent Space agents can autonomously analyze enterprise data (like portfolio information from Salesforce, OneDrive, and other connected sources), identify risks and opportunities, and generate actionable summaries without manual intervention, such as flagging cash flow problems in client accounts.
“Agent Space has identified a cash flow problem with Acme General Contracting... Agent Space has already identified that possibility as a suggested follow-up. So now let's go ahead and deep dive into general contracting industry trends. This activates Google's enterprise deep research agent, which starts by telling me what it plans to research and in what order.”
Agent Space provides an ecosystem where businesses and developers can create agents that work seamlessly together, powered by an authenticated, personalized interface that allows approved agents from multiple sources (Google-made, company-created, partner-built, and custom user-built agents) to coexist and be accessed from a single dashboard.
“The agent gallery lets me see my company's approved selection of purpose-built agents, including ones powered by third party models like Llama and Claude. You see, we've got some Google made agents. We have agents that my bank has made available to me, either ones we've created or ones built by partners. And then the best part, my own personal agents, which I can build directly inside agent space with this button over here, or even easier, just from having a little conversation.”
The A2A protocol provides enterprise-grade authentication and authorization support, enabling secure collaboration between agents while maintaining fine-grained access controls required by regulated industries.
“secure by default with enterprisegrade authentication and authorization support for long running tasks.”
Agent Space provides stringent access controls at the employee level and can operate within a customer's own VPC, ensuring data privacy and compliance with regulatory requirements for regulated industries like banking.
“For companies with strict regulatory needs like a bank, agent, agentspace provides stringent access controls at the employee level and can operate within your own VPC, ensuring that your data stays yours while meeting all of your requirements.”
The A2A protocol is built on existing standards to enable easy integration with existing IT stacks and business infrastructure, reducing migration costs and enabling incremental adoption across organizations.
“It's built on a lot of the existing standards. So it's easy to integrate it with existing tech with existing IT stacks or businesses already use.”
The A2A protocol supports long-running tasks that may span hours or days, with optional human-in-the-loop workflows, enabling complex business processes that require asynchronous coordination and human oversight.
“Everything from short quick tasks to deep research that might take hours or even days when humans are in the loop. So in that example we saw that that took days right two days later you have the people interviewed and then you know two business days were waiting for them to get their background checks etc.”
In agent-to-agent communication, a client agent and remote agent can negotiate interaction modalities (text, forms, audio, video) and maintain state management, ensuring they can communicate in mutually compatible formats while exchanging task context and updates.
“Next, you can describe to the agent how you would like to communicate using text, forms, audio, video, etc... Next, the client and the agent kind of talk back and forth. The agents are able to ask for clarifications information and do various sub actions on behalf of the clients or users.”
Agent Space can analyze enterprise data from sources like OneDrive, Salesforce, or Databricks, interpreting user queries to surface relevant information, identifying risks and opportunities (such as a client's potential cash flow problems), and generating reports with source links that users can then act upon.
“It knows which clients are mine and summarizes top points from my data sources like one drive, Salesforce or done in Bradre's. If I have questions, I have a direct link to my sources here... Agent Space doesn't just summarize information. It's interpreting my question and surfacing what matters most. For example, in this chart, I can see Agent Space has flagged that Acme General Contracting might have some cash flow problems in the future. Already, it's given me a massive report.”
Agent Space is part of Google Cloud Next 2025 announcements, includes an Agent Development Kit (ADK) which is an open-source flexible framework for developing and deploying multi-agent systems, and has over 50 technology partners including Atlassian, Salesforce, Deloitte, Langchain, Anthropic, Inuit, Jet Brains, MongoDB, PayPal, and others.
“Here's the kind of the agentic ecosystem or at least the the partnerships that are providing various agents and various capabilities. We have big names Atlation, Salesforce, Deote, Langchain and Inuit, Jet Brains, many many more... They introduced it at the current conference, the Google Cloud Next 2025, and it empowers developers to build production ready agentic applications.”
Anthropic's Model Context Protocol (MCP) allows large language models to communicate with other software; the A2A protocol complements MCP by enabling agent-to-agent communication at a higher level of abstraction than model-to-system communication.
“And you might have heard about Anthropics MCP, the model context protocol, which allows these large language models to communicate with a lot of other software... And this A2A is an open protocol that complements the MCP.”
In the A2A protocol architecture, the client agent represents the user and is responsible for secure collaboration, task and state management, user experience negotiation, and capability discovery with remote agents, while the remote agent performs the delegated work and communicates results back.
“So you have your client agent representing you. You have the remote agent and your client agent does the secure collaboration task and state management, user experience negotiation, capability discovery. So we kind of covered that in the video. And so this protocol, what it does is it communicates between the client agent and a remote agent. Right? So the client agent is the thing that kind of represents you. It's it's your people, the people working for you. the agents working for you and then it works with the remote agent giving it whatever context it needs and working with it updating it etc to complete the task that you you set for it to be carried out.”
Agent Space allows users to request agents to draft communications (e.g., emails) based on analysis and insights, and users can send these directly from within Agent Space without switching to external email applications, with sessions saved for later continuation.
“I need them to see it right away so I can ask agent space. Draft me an email to Acme General Contracting CEO requesting a meeting for next week. And just like that, I've got the draft ready to go. And even better, I can send it off directly from within Agent Space. So I don't even have to switch to Outlook or Gmail. I'm all set. And Agent Space has saved my session, so I can prep for that meeting right where I left off whenever I'm ready.”
Agent Space's approach to data privacy and security includes restricting agents' data access to information the user has explicit permission to access, with agents aware of their own data scope (e.g., knowing which clients they are authorized to see).
“This only uses information that I have permission to access. It knows which clients are mine and summarizes top points from my data sources like one drive, Salesforce or done in Bradre's.”
The A2A protocol is modality-agnostic, supporting communication through text, forms, audio, video, and streaming, allowing agents to interact using the most appropriate communication medium for each task.
“It's modality agnostic so it's not just text it can do whatever you want audio video streaming etc.”
The two agents in the A2A protocol communicate with task completion as the primary orientation, using structured 'artifacts' to pass outputs between agents, creating a work-focused communication pattern rather than open-ended conversation.
“The two agents communicate and all of their sort of communication is oriented towards task completion. It seems like they're calling the output of a task uh an artifact.”
Google's historical success came from creating a search-bar gateway (Google.com) that directed users to relevant websites, monetized through advertisements; as AI adoption grows, Google is pivoting from search to agents, where instead of a search bar, users deploy personal agents that discover and select service providers, enabling Google to monetize through agents paying for preferential placement.
“if you think about what made Google so successful back in the days is they basically created a search bar as sort of the gate to the web. You type in what you want and it takes you to where you need to go. It was very different from how other people build their businesses online because instead of trying to keep them in their sort of website on google.com, it was trying to get them off of google.com, right? So you type in background checks for employment. For most of Google's history, they wanted you off of Google and on to, for example, this ad or this ad or this ad. In the future, of course, it's probably going to go to instead of where we're typing in something to a search bar, we're just going to ask our agent to do it. It's going to be some sort of a maybe it's a chatbot or maybe at some point we're going to have more of like a video, some sort of an AI avatar, audio, video, whatever. But the point is for some of these services, these businesses might want to pay to have their agent appear above the other agents that are available. How they do that? probably by paying Google some money to have it prioritize their services, their agent.”
Agent Space integrates time series forecasting models specifically trained for financial scenarios, enabling cash flow predictions across multiple quarters with confidence estimates and actionable recommendations.
“I'm going to have our bank's cash flow agent do some forecasting across the next three quarters for me. This agent uses Google's new time series forecasting model, which is specifically trained for scenarios just like this.”
The speaker uses a hiring and background check scenario to illustrate the practical implementation of the A2A protocol, demonstrating how agents handle capability constraints (domestic vs. international) and discover alternative agents to complete tasks.
“For example, here we're asking this agent, I am hiring for a staff software engineer position. Attached is the job description. Can you help me fill this role? The agent goes on to uh source candidates and looks through the various available agents for one that's able to do candidate sourcing”
Users can build custom agents in Agent Space without writing code by simply having a conversation with the agent, which automatically generates an agent plan and enables refinement through natural language dialogue.
“Agent Space automatically generates an agent plan based on our conversation so far. And this is good, but I think I want more. I'm gonna have it generate an audio summary and send it to my inbox so I can listen to it on my morning commute. And just like that, I have built my own custom agent to use whenever I want without writing a single line of code.”
The A2A protocol uses a simple JSON format to allow agents to advertise their capabilities, enabling standardized capability discovery and allowing service providers to charge fees for agent services based on usage.
“It looks like these agents they can advertise their capability. It's using a simple JSON format. So you kind of list the capabilities like we saw for example US-based background checks right you can say that's one of the capabilities that your agent can do.”
As the A2A agent ecosystem grows with more capable agents added to it, it will become easy to monetize the platform through a single chat interface that provides access to millions of agents, with any agent capable of performing a task being plugged in and made available to users.
“This seems like it could be big, massive. I mean, if you think about it, if this ecosystem grows, more and more capable agents are added to it, it'll be very easy to to monetize it. For example, it'll be very easy to use it. Basically, one chat window, one chat bot could give you an access to millions of agents, perhaps, however many will exist on this platform. So basically any agent that wants to be plugged in here could be.”
The historical success of Google as a search engine relied on creating a gateway (the search bar at Google.com) that users accessed to find information, and Google monetized this by placing advertisements, but as AI adoption grows, the traditional search bar entry point may become outdated because users increasingly interact with AI agents and chatbots that handle information retrieval on their behalf.
“The reason why this could be a big big deal is because just like back in the days, Google dominated sort of like the web space by just having a search bar, right? Google.com. You type in what you want and it goes and it finds whatever website you need. Made all its money by putting some ads here and there. As AI takes over, that might be a little bit outdated. The idea of search. Now more and more people familiar with AI, we find ourselves using a chatbot, some sort of an AI agent that does tasks on our behalf that finds information for us. The entry to the web isn't necessarily the search bar. It might be your sort of agent, your client, your chatbot that represents you and then goes out there and finds the information for you.”
Google's Agent Space ecosystem fundamentally shifts the entry point to digital services from search bars to personal AI agents that represent users and find information on their behalf, potentially replicating Google's original dominance model in an agentic future.
“The idea of search. Now more and more people familiar with AI, we find ourselves using a chatbot, some sort of an AI agent that does tasks on our behalf that finds information for us. The entry to the web isn't necessarily the search bar. It might be your sort of agent, your client, your chatbot that represents you and then goes out there and finds the information for you.”
More than 50 technology partners are participating in the A2A protocol, including major names such as Atlassian, Salesforce, Deloitte, Langchain, MongoDB, PayPal, Inuit, and Jet Brains, indicating broad industry adoption and commitment to the open standard.
“They have more than 50 technology partners aboard. So we're talking about Atlation, Coher into it, Lankchain, MongoDB, PayPal, Salesforce, tons of names that you would recognize, Deote, etc.”
The A2A (agent-to-agent) protocol is an open protocol that enables agents to work together, securely exchange information, and coordinate actions even when built by different vendors or frameworks, allowing for increased autonomy and multiplied productivity gains while lowering long-term costs.
“enabling agents to interoperate with each other even if they were built by different vendors or in a different framework will increase autonomy and multiply productivity gains while lowering long-term costs. And hence they're launching this new open protocol called A2A agent to agent.”
In future interactions, people may increasingly ask 'Have your agents talked to my agents?'—suggesting that agent-to-agent communication will become commonplace in business and personal workflows.
“In the future, I think we'll just all be saying, "Have your agents talked to my agents?"”
If the A2A agent ecosystem grows with an increasing number of capable agents, it becomes easy to monetize by having users access millions of agents through a single chat interface or chatbot, creating network effects where platform value increases exponentially as agent inventory grows.
“If you think about it, if this ecosystem grows, more and more capable agents are added to it, it'll be very easy to to monetize it. For example, it'll be very easy to use it. Basically, one chat window, one chat bot could give you an access to millions of agents, perhaps, however many will exist on this platform.”
Google's competitive advantage in an agent-mediated future depends on creating a platform where service providers will pay Google for preferential placement of their agents, similar to its historical ad-based model where search results were monetized through advertiser payments.
“In the future, of course, it's probably going to go to instead of where we're typing in something to a search bar, we're just going to ask our agent to do it... the point is for some of these services, these businesses might want to pay to have their agent appear above the other agents that are available. How they do that? probably by paying Google some money to have it prioritize their services, their agent. This might be Google 2.0.”
A single client agent can potentially provide access to millions of available service agents through a unified interface, creating a convergence point for diverse services similar to Google's historical role as the gateway to the web.
“One chat window, one chat bot could give you an access to millions of agents, perhaps, however many will exist on this platform. So basically any agent that wants to be plugged in here could be.”
The A2A protocol monetization model could enable service providers to charge fees for their agent services, with agents sitting in the registry and advertising their services, and when clients need those services, the client agent interfaces with the remote agent to complete the task and potentially enabling automated payment handling.
“And this is what I mean by it can be easily monetized because if you're a company that's providing those services you might charge a fee of some sort for that for doing that work but your agent kind of sit there and advertise its services like I am able to do background checks within the US and then whenever somebody needs that right that little agent interfaces the remote agent interfaces with the client agent to complete that. I haven't seen any reference to how payments get handled yet but obviously that's going to come in at some point. I mean, there's going to be some sort of a credit system, some sort of a point system, whatever, that you can buy certain amounts of credits. I'm just guessing, but maybe it's just the agent being able to like charge your credit card.”
The Agent Development Kit (ADK) is an open-source, flexible framework designed to simplify the development and deployment of multi-agent systems, making it easier for developers to build and integrate agents into the broader A2A ecosystem.
“Lavi, a machine learning engineer at Google. So, he posted a couple things. For example, they're sharing the agent development kit, ADK. It's an open-source flexible framework designed to simplify the development and deployment of multi-agent systems.”
Google's strategy with Agent Space represents a fundamental business model evolution that could be termed 'Google 2.0,' where the company maintains its position as a digital gatekeeper by shifting the primary interface from search queries to agentic delegation.
“This might be Google 2.0. We'll see. But it certainly seems like this is where they're trying to head to now.”
Agent Space might represent 'Google 2.0'—Google's strategic evolution from being the search gateway to the web to becoming the orchestration platform for an agent-mediated economy.
“This might be Google 2.0. We'll see. But it certainly seems like this is where they're trying to head to now.”
With A2A protocol adoption, users can combine agents from various providers and have them work together seamlessly in a plug-and-play manner, allowing users to select and orchestrate agents based on capability matching without concern for the agent's vendor or underlying framework.
“the big point here is that as a user, you're allowed to sort of combine various agents from various providers, kind of put them together and have them just go and do whatever task you need them to do seamlessly. You just like pick and choose. It doesn't matter what company they're from. It's plugandplay. As long as the agent is capable of doing that task, you can easily call it to work on your behalf.”
The A2A protocol enables users to combine agents from different vendors and providers, selecting them based on task requirements without vendor lock-in, creating a plug-and-play marketplace where any capable agent can be invoked for its specific service.
“The big point here is that as a user, you're allowed to sort of combine various agents from various providers, kind of put them together and have them just go and do whatever task you need them to do seamlessly. You just like pick and choose. It doesn't matter what company they're from. It's plugandplay. As long as the agent is capable of doing that task, you can easily call it to work on your behalf on whatever project you're doing.”
If agents can advertise their capabilities through the A2A protocol, service providers can monetize by charging fees for agent services, with potential payment mechanisms including credit systems, point systems, or direct credit card charges to users.
“This is what I mean by it can be easily monetized because if you're a company that's providing those services you might charge a fee of some sort for that for doing that work but your agent kind of sit there and advertise its services like I am able to do background checks within the US and then whenever somebody needs that right that little agent interfaces the remote agent interfaces with the client agent to complete that. I haven't seen any reference to how payments get handled yet but obviously that's going to come in at some point. I mean, there's going to be some sort of a credit system, some sort of a point system, whatever, that you can buy certain amounts of credits. I'm just guessing, but maybe it's just the agent being able to like charge your credit card.”
The A2A protocol is complementary to Anthropic's Model Context Protocol (MCP), which allows large language models to communicate with external software, indicating that multiple interoperability standards are emerging alongside the agent ecosystem.
“You might have heard about Anthropics MCP, the model context protocol, which allows these large language models to communicate with a lot of other software. We saw one that got integrated into Blender and was basically using Claude to create 3D scenes in Blender. And this A2A is an open protocol that complements the MCP.”
Agent Space is powered by Gemini 2.5 and Google Search technology, and it is the only hyperscaler platform on the market that can connect to third-party data and tools while offering interoperability with third-party agents and models.
“Powered by Gemini 2.5 and Google Search technology, Agent Space is the only hyperscaler platform on the market that can connect to thirdparty data and tools and offers interoperability with thirdparty agents and models.”
Agent Space is currently available by waitlist; users need to sign up providing name, business email, phone number, and job role, and based on other Google product launches, there will likely be a long waiting list due to high demand.
“Word to the wise, you better sign up for this ASAP cuz there's going to be probably a pretty long waiting list. if the other things that Google just announced are any indication that a lot of people are going to be pouring into this thing trying to access it and use it. So, search for Google agent space. Click on sign up for access. They just want your name, business email, phone number, job roll.”
Agent Space is powered by Gemini 2.5 and Google Search technology, and is positioned as the only hyperscaler platform that can connect to third-party data and tools while offering interoperability with third-party agents and models.
“Powered by Gemini 2.5 and Google Search technology, Agent Space is the only hyperscaler platform on the market that can connect to thirdparty data and tools and offers interoperability with thirdparty agents and models.”
Agent Space allows agents to generate drafts of business communications (such as emails to clients) and send them directly within the platform without requiring the user to switch to external applications like Outlook or Gmail.
“Draft me an email to Acme General Contracting CEO requesting a meeting for next week. And just like that, I've got the draft ready to go. And even better, I can send it off directly from within Agent Space. So I don't even have to switch to Outlook or Gmail.”
Agent Space saves user sessions and conversation history, allowing users to return to unfinished tasks at the exact point where they left off without needing to restart their workflow.
“Agent Space has saved my session, so I can prep for that meeting right where I left off whenever I'm ready.”
The Agent Development Kit (ADK) was introduced at Google Cloud Next 2025 and empowers developers to build production-ready agentic applications with greater flexibility and precise control over multi-agent system behavior.
“They introduced it at the current conference, the Google Cloud Next 2025, and it empowers developers to build production ready agentic applications with greater flexibility and precise control.”
The proliferation of AI and ML-specific acronyms (MCP, A2A, ADK, etc.) represents a complexity management challenge for users adopting these technologies, with the number of acronyms likely to continue growing as AI systems become more sophisticated.
“I feel like a lot of these acronyms will make things very hard to follow. They're especially common in AI and ML. So, you've probably memorized the acronym MCP and now add one more to it. A2A. Just imagine how many more we'll have to memorize before we get to AGI.”
The speaker plans to conduct a full test of Agent Space and share the results once they gain access to the platform, providing empirical evaluation of the product's capabilities.
“We'll do a full test of this thing. The second I get my hands on it”
The speaker intends to conduct a full test of Agent Space once they gain access and plans to provide a detailed evaluation.
“We'll do a full test of this thing. The second I get my hands on it”
Agent Space access is currently limited to users who sign up for a waitlist and receive approval, and there will likely be a long waiting list given the level of interest in Google's recent AI announcements.
“Word to the wise, you better sign up for this ASAP cuz there's going to be probably a pretty long waiting list. if the other things that Google just announced are any indication that a lot of people are going to be pouring into this thing trying to access it and use it.”
Users interested in Agent Space can sign up through Google's website by searching 'Google agent space' and clicking a signup link, providing basic information (name, business email, phone number, job role) across approximately seven form fields.
“So, search for Google agent space. Click on sign up for access. They just want your name, business email, phone number, job roll. There's like seven fields to fill out and you're going to be on the wait list.”