Why Your AI Agents Can't Talk to Each Other (Yet) — Vlad Luzin, BAND

OK. Hello everyone. Can OK. Hello everyone. Can you hear me? Many you hear me? Many you hear me? Many people, yes. So, people, yes. So, people, yes. So, please, be quiet. My please, be quiet. My please, be quiet. My name is Vlad. I am the name is Vlad. I am the name is Vlad. I am the co-founder and co-founder and co-founder and CTO at CTO at CTO at Bend. Today I want to Bend. Today I want to Bend. Today I want to discuss discuss discuss four topics with you. First, the four topics with you. First, the four topics with you. First, the thesis, our thesis, our thesis, our company's thesis, what we company's thesis, what we company's thesis, what we believe in. The evolution of AI: believe in. The evolution of AI: believe in. The evolution of AI: from adversarial from adversarial from adversarial agents to agents to agents to cycle engineering and beyond. cycle engineering and beyond. Technical challenges that Technical challenges that we are currently experiencing and we are currently experiencing and we are currently experiencing and that should be addressed that should be addressed that should be addressed immediately. And, of course, Bend immediately. And, of course, Bend immediately. And, of course, Bend to introduce to introduce to introduce the company, the product, and what the company, the product, and what the company, the product, and what we do. we do. we do. Let's start with the thesis. First Let's start with the thesis. First Let's start with the thesis. First of all, we believe that the of all, we believe that the of all, we believe that the future lies in AI-to- future lies in AI-to- future lies in AI-to- AI communication AI communication AI communication within businesses, within businesses, within businesses, between businesses, and between businesses, and between businesses, and between between between consumers and consumers and consumers and businesses. AI businesses. AI businesses. AI will do the work will do the work will do the work on our behalf, and they on our behalf, and they on our behalf, and they will have to will have to will have to communicate with each communicate with each communicate with each other to solve other to solve other to solve tasks for us. tasks for us. Second, agents will be Second, agents will be Second, agents will be autonomous, autonomous, autonomous, constantly operating constantly operating constantly operating entities entities entities distributed around the distributed around the distributed around the world. These are world. These are world. These are nice words, but I nice words, but I nice words, but I want to take want to take want to take a minute to let you a minute to let you a minute to let you imagine what imagine what imagine what this this this type of communication between type of communication between type of communication between agents would actually look like. Demo time.

agents would actually look like. Demo time. Agents will Agents will communicate in a communicate in a communicate in a conversational space. conversational space. conversational space. They will see each They will see each They will see each other. They will be other. They will be other. They will be distributed. They distributed. They distributed. They will receive will receive will receive tasks from us or tasks from us or tasks from us or from any system from any system . They will be able to search for . They will be able to search for . They will be able to search for colleagues in registries, colleagues in registries, colleagues in registries, invite them to the invite them to the invite them to the conversation space, conversation space, conversation space, delegate tasks, delegate tasks, delegate tasks, collect information, collect information, collect information, and get back to us and get back to us . This is what . This is what . This is what the future of the future of the future of AI-to-AI communication will look like. So, AI-to-AI communication will look like. So, AI-to-AI communication will look like. So, keep that keep that keep that in mind as we speak. Let's in mind as we speak. Let's in mind as we speak. Let's talk a little about talk a little about talk a little about competitive agents. competitive agents. I think everyone here knows I think everyone here knows this concept, and I'm this concept, and I'm this concept, and I'm sure everyone here sure everyone here uses it too. Where do we uses it too. Where do we use it? use it? use it? If you If you If you use Claude use Claude use Claude or Codex, I'm sure or Codex, I'm sure or Codex, I'm sure you run you run you run multiple sessions to multiple sessions to multiple sessions to work on the same work on the same work on the same task. One task. One task. One plans, the other plans, the other plans, the other checks. One writes checks. One writes checks. One writes the code, the other checks the the code, the other checks the the code, the other checks the code. So, essentially, code. So, essentially, code. So, essentially, you have two you have two you have two sexual agents working on the sexual agents working on the sexual agents working on the same task, and you are same task, and you are like a Cisco like a Cisco router or router or router or switch that switch that switch that forwards packets between forwards packets between forwards packets between them. Let's talk about them. Let's talk about them. Let's talk about cycle engineering. There cycle engineering. There cycle engineering. There were were were a lot of talks about a lot of talks about a lot of talks about cycle engineering at this conference, cycle engineering at this conference, cycle engineering at this conference, a lot of articles, it's a a lot of articles, it's a a lot of articles, it's a pretty new pretty new pretty new concept, right?

concept, right? concept, right? But I would like to get it down But I would like to get it down But I would like to get it down to the point. So, what to the point. So, what to the point. So, what is is is cycle engineering? In principle, this is the cycle engineering? In principle, this is the cycle engineering? In principle, this is the same as the same as the same as the previous concept. previous concept. You're running You're running multiple agents, multiple agents, multiple agents, preferably with preferably with preferably with stateful support rather than without, stateful support rather than without, stateful support rather than without, but instead of but instead of but instead of being a router being a router being a router or a switch, you or a switch, you or a switch, you delegate that to some delegate that to some delegate that to some Python or TypeScript code Python or TypeScript code Python or TypeScript code written by someone as an written by someone as an written by someone as an open-source project, and that open-source project, and that open-source project, and that code just sends code just sends code just sends requests back requests back requests back and forth to the agents, and you're and forth to the agents, and you're and forth to the agents, and you're just running the just running the just running the process. So why do we process. So why do we process. So why do we use these use these use these agents when agents when agents when we do this kind of we do this kind of we do this kind of work? Because of the work? Because of the work? Because of the architectural architectural architectural limitations of limitations of limitations of transformers, namely transformers, namely transformers, namely because of a concept because of a concept because of a concept called the " called the " called the " single agent bottleneck single agent bottleneck ", right? So, why don't we ", right? So, why don't we ", right? So, why don't we run one run one run one agent in one session agent in one session agent in one session that does everything? that does everything? that does everything? For example, For example, For example, the tendency to the tendency to the tendency to confirm one confirm one confirm one 's own thoughts, right? 's own thoughts, right? 's own thoughts, right? Blurred attention, Blurred attention, Blurred attention, fragmentation fragmentation fragmentation of context, of context, of context, impaired impaired impaired recall of recall of recall of information. So a information. So a information. So a context of 1 or 2 context of 1 or 2 context of 1 or 2 million tokens will not million tokens will not million tokens will not help you help you help you improve improve improve performance. But performance. But performance. But let's talk let's talk let's talk about multi-agent about multi-agent about multi-agent systems, okay? We systems, okay? We systems, okay? We already understand that already understand that already understand that launching multiple sessions launching multiple sessions is, in essence, the work of is, in essence, the work of is, in essence, the work of multi-agent multi-agent multi-agent systems. What is the systems. What is the systems. What is the easiest way easiest way easiest way to combine the work of to combine the work of to combine the work of multiple agents multiple agents multiple agents together? Messengers, together? Messengers, together? Messengers, right? We already have Slack, right? We already have Slack, right? We already have Slack, right? Anthropic right? Anthropic right? Anthropic released an agent for released an agent for released an agent for Slack, we have Teams, Discord,

Slack, we have Teams, Discord, Slack, we have Teams, Discord, WhatsApp, Telegram, etc., WhatsApp, Telegram, etc., WhatsApp, Telegram, etc., right? So, this is the right? So, this is the right? So, this is the solution. Well, to solution. Well, to solution. Well, to connect an agent to connect an agent to connect an agent to Telegram, it takes five Telegram, it takes five Telegram, it takes five steps. There are steps. There are seven steps to connecting to Discord. There are eight eight steps to connecting an agent to Slack. For WhatsApp—11. And steps to connecting an agent to Slack. For WhatsApp—11. And steps to connecting an agent to Slack. For WhatsApp—11. And these are not simple steps at all these are not simple steps at all these are not simple steps at all . They are all . They are all . They are all done manually and done manually and done manually and you have to read a you have to read a you have to read a whole book to whole book to whole book to set it up. And even set it up. And even set it up. And even if you do, you will if you do, you will if you do, you will only get one thing, only get one thing, only get one thing, and nothing more. What is this and nothing more. What is this and nothing more. What is this ? An agent ? An agent ? An agent communicating with communicating with communicating with a person, usually a person, usually a person, usually you. This agent can't you. This agent can't you. This agent can't talk to talk to talk to other agents, can't other agents, can't other agents, can't talk to talk to talk to other people until other people until other people until you go through the you go through the you go through the seven circles of hell. So, seven circles of hell. So, seven circles of hell. So, essentially, your agents essentially, your agents essentially, your agents are still in a are still in a are still in a kind of digital kind of digital kind of digital isolation. But smart isolation. But smart isolation. But smart people will say, “I know the people will say, “I know the people will say, “I know the solution. Protocols, solution. Protocols, solution. Protocols, right? MCP, ACP, A2A, etc. This right? MCP, ACP, A2A, etc. This right? MCP, ACP, A2A, etc. This is the solution." Isn't that right?

is the solution." Isn't that right? Well, if you are Well, if you are a fan of this a fan of this a fan of this approach to approach to approach to connecting your connecting your connecting your sessions, then this is what sessions, then this is what sessions, then this is what your your your multi-agent multi-agent multi-agent system will look like. Several system will look like. Several system will look like. Several servers call servers call servers call other agents as " other agents as " tools" through MCP tools" through MCP . And if you want, . And if you want, . And if you want, maybe this agent that maybe this agent that maybe this agent that you called as a you called as a you called as a tool can tool can tool can call another call another call another agent through A2A. agent through A2A. agent through A2A. It sounds extremely It sounds extremely It sounds extremely simple. Okay, everyone simple. Okay, everyone simple. Okay, everyone knows what A2A is. knows what A2A is. knows what A2A is. Google has a very huge Google has a very huge Google has a very huge marketing budget marketing budget . But let's . But let's . But let's look at the look at the look at the difficulties. MCP stands for difficulties. MCP stands for stateless calls. This stateless calls. This means you can't means you can't means you can't go back go back go back to the agent you to the agent you to the agent you spoke to and spoke to and spoke to and ask again what happened. ask again what happened. ask again what happened. A2A communication is A2A communication is A2A communication is one-way unless one-way unless one-way unless you implement you implement you implement both directions. So both directions. So , agent A sends , agent A sends , agent A sends a task to agent B. This is a a task to agent B. This is a a task to agent B. This is a client-server. If client-server. If client-server. If you want agent B you want agent B you want agent B to send a task to to send a task to to send a task to agent A, that's also agent A, that's also agent A, that's also client-server. Therefore, client-server. Therefore, client-server. Therefore, you need to you need to you need to implement both implement both implement both the client and the the client and the the client and the server on both sides. Chaining server on both sides. Chaining server on both sides. Chaining REST API calls REST API calls REST API calls will result in will result in timeouts. And of course, there are timeouts. And of course, there are timeouts. And of course, there are many many many things missing, like things missing, like things missing, like discovery, discovery, discovery, A2A session state, queues, which A2A session state, queues, which A2A session state, queues, which you still need to you still need to you still need to implement and add implement and add implement and add to all of this. And to all of this. And to all of this. And if you do that, you are if you do that, you are if you do that, you are actually building a actually building a actually building a layout. You are not layout. You are not layout. You are not building a building a building a multi-agent multi-agent multi-agent system. You system. You system. You waste time on waste time on waste time on boring, dull things.

boring, dull things. Now let's summarize what Now let's summarize what we know. We we know. We we know. We know that know that know that multi-agent multi-agent multi-agent systems are already here. systems are already here. systems are already here. Everyone here Everyone here Everyone here uses multiple uses multiple uses multiple agents. We know that agents. We know that agents. We know that messaging platforms messaging platforms messaging platforms are are are not the solution. They are not the solution. They are not the solution. They are designed for humans and designed for humans and designed for humans and actively block actively block bot-to-bot attempts. And we bot-to-bot attempts. And we know that a protocol know that a protocol is too is too is too low-level a low-level a low-level a technical abstraction technical abstraction technical abstraction to to to build anything on at build anything on at build anything on at scale. But scale. But scale. But how difficult is it how difficult is it how difficult is it to create a system that to create a system that to create a system that allows multiple allows multiple allows multiple agents agents agents to connect together? to connect together? to connect together? It's not that It's not that It's not that difficult, is it? difficult, is it? You see, connecting You see, connecting remote agents remote agents remote agents to sessions as a process to sessions as a process to sessions as a process running on your running on your running on your laptop or my laptop or my laptop or my cloud to your cloud to your cloud to your code—that's a code—that's a code—that's a distributed systems problem. distributed systems problem. distributed systems problem. And distributed And distributed And distributed systems systems systems are by definition not are by definition not are by definition not simple. Forget simple. Forget simple. Forget about agents, okay? about agents, okay? about agents, okay? Deterministic Deterministic software is a software is a headache.

headache. headache. Add to that Add to that Add to that agents that are agents that are agents that are microservices and microservices and microservices and non-deterministic, non-deterministic, non-deterministic, okay? And you will get a okay? And you will get a okay? And you will get a lot of "satisfaction" lot of "satisfaction" . To solve this, you . To solve this, you . To solve this, you actually need to actually need to actually need to pay attention to the pay attention to the pay attention to the transport layer, transport layer, transport layer, right? Transportation right? Transportation right? Transportation should be in real should be in real should be in real time. It must be time. It must be time. It must be ordered ordered ordered because LLMs expect because LLMs expect because LLMs expect ordered ordered ordered messages. You messages. You messages. You need to handle need to handle need to handle retries, etc. retries, etc. retries, etc. You need to You need to You need to handle handle handle persistence and persistence and persistence and hydration because hydration because hydration because your agent is a your agent is a your agent is a microservice. It could microservice. It could microservice. It could fail, the pod could fail, the pod could fail, the pod could crash, and then crash, and then crash, and then your entire multi-agent your entire multi-agent your entire multi-agent system would fail. system would fail. system would fail. So, you need to So, you need to So, you need to take that into account. You take that into account. You take that into account. You need to deal with need to deal with need to deal with something very new— something very new— runtime binding. runtime binding. If you want to If you want to If you want to connect LangGraph to connect LangGraph to connect LangGraph to the code or the cloud to the code or the cloud to the code or the cloud to CrewAI, you need to CrewAI, you need to CrewAI, you need to connect the connect the connect the thread ID thread ID thread ID to the to the session ID, execution ID, and launch ID of these multiple launch ID of these multiple agents and map agents and map agents and map these IDs these IDs these IDs to each other. Otherwise, to each other. Otherwise, to each other. Otherwise, agents will not be able to agents will not be able to agents will not be able to work as a group work as a group work as a group on a single goal or on a single goal or on a single goal or task. And of course, task. And of course, task. And of course, you can't you can't you can't work at the level of IP work at the level of IP addresses, ports, addresses, ports, URLs, or even pub URLs, or even pub /sub topics. You need to /sub topics. You need to /sub topics. You need to rise above all the rise above all the rise above all the technical details technical details technical details to another level of to another level of to another level of abstraction: conversations, abstraction: conversations, abstraction: conversations, participants, channels or participants, channels or participants, channels or rooms, and rooms, and message routing message routing adapted for adapted for adapted for agents as the main agents as the main agents as the main entities. And, of course, all of this is still entities. And, of course, all of this is still not enough to not enough to use such a use such a use such a product or product or product or communication software,

communication software, communication software, because every organization because every organization because every organization seeks control. They seeks control. They seeks control. They need need need identification and identification and identification and observability. observability. observability. Observability is not Observability is not Observability is not just about the message just about the message just about the message sent from sent from sent from agent A on server A agent A on server A agent A on server A to agent B on to agent B on to agent B on server B. It's also about server B. It's also about server B. It's also about what tool calls what tool calls occurred after the occurred after the agent received that agent received that agent received that message, right? And it is message, right? And it is message, right? And it is very difficult very difficult very difficult to provide this to provide this to provide this level of level of level of observability in observability in observability in distributed distributed distributed systems. So without systems. So without systems. So without addressing all these addressing all these addressing all these issues, agents won't issues, agents won't issues, agents won't be able to communicate be able to communicate be able to communicate with each other, and this with each other, and this with each other, and this beautiful future of ours, beautiful future of ours, beautiful future of ours, where everyone gets a where everyone gets a where everyone gets a basic income and basic income and basic income and lies on the beach, won't lies on the beach, won't lies on the beach, won't come. I want to come. I want to come. I want to introduce Bend. We introduce Bend. We introduce Bend. We actually took everything we actually took everything we actually took everything we talked about and talked about and talked about and resolved it. So, you resolved it. So, you resolved it. So, you don't have to don't have to don't have to solve this on solve this on solve this on your own. You can your own. You can your own. You can very easily very easily very easily connect all connect all connect all your agents within the your agents within the your agents within the enterprise or enterprise or enterprise or your Hermes and Open Cloak. I your Hermes and Open Cloak. I your Hermes and Open Cloak. I want to introduce a want to introduce a want to introduce a product we are product we are product we are launching this launching this launching this week. This is a global week. This is a global week. This is a global level of interaction and level of interaction and level of interaction and collaboration for collaboration for any agents, any agents, any platform, language, any platform, language, any platform, language, or environment. And it's not or environment. And it's not or environment. And it's not just peer-to-peer, it's just peer-to-peer, it's just peer-to-peer, it's multi-peer. And we haven't multi-peer. And we haven't multi-peer. And we haven't forgotten about people either.

forgotten about people either. forgotten about people either. If you are a human, you If you are a human, you If you are a human, you can still can still can still connect and connect and connect and talk to your talk to your talk to your agent. This is completely agent. This is completely agent. This is completely normal. And if we normal. And if we normal. And if we look under the hood, look under the hood, look under the hood, we will see that it is not we will see that it is not we will see that it is not just a just a just a communication communication communication layer. We have layer. We have layer. We have a registry, so your a registry, so your a registry, so your agents see each agents see each agents see each other automatically, other automatically, other automatically, without any extra effort on without any extra effort on without any extra effort on your part. your part. your part. Persistence, Persistence, Persistence, channels, channels, channels, message filtering. Instead of message filtering. Instead of message filtering. Instead of doing it on the doing it on the doing it on the client side, we client side, we client side, we do it inside the do it inside the do it inside the platform. Security, platform. Security, platform. Security, observability, observability, observability, etc. Now the etc. Now the etc. Now the most interesting thing. most interesting thing. most interesting thing. Let me show Let me show Let me show you some demos. Time is you some demos. Time is you some demos. Time is short, so the demos are short, so the demos are short, so the demos are recorded, but we have a recorded, but we have a recorded, but we have a booth. Please booth. Please booth. Please come watch come watch come watch live or just live or just live or just tune in. The demo tune in. The demo tune in. The demo will show two will show two will show two users in two users in two users in two different browsers different browsers different browsers connected to the connected to the connected to the communication communication communication platform. One is me, platform. One is me, platform. One is me, Vlad. I have a Vlad. I have a Vlad. I have a personal agent, personal agent, personal agent, Claude, who works on a Mac. And the Claude, who works on a Mac. And the Claude, who works on a Mac. And the other user— other user— Mike, with Claude code in the Mike, with Claude code in the Mike, with Claude code in the terminal session, okay? Um, terminal session, okay? Um, terminal session, okay? Um, Claude SDK and um Landlord. Claude SDK and um Landlord. Demonstration number Demonstration number one. In the top one. In the top one. In the top window—me, my account.

window—me, my account. window—me, my account. And I have a And I have a And I have a personal personal personal assistant. Below is assistant. Below is another user another user another user connected to the same connected to the same connected to the same interaction layer, without interaction layer, without interaction layer, without any agents. any agents. Right, okay, I'll Right, okay, I'll launch two agents, launch two agents, launch two agents, they'll join the they'll join the they'll join the platform and become platform and become platform and become agents owned by agents owned by agents owned by Mike. And let's see Mike. And let's see Mike. And let's see how fast it will how fast it will how fast it will be. So, first be. So, first be. So, first we start the agent. we start the agent. Within seconds, the Within seconds, the agent card agent card agent card appears in appears in appears in Mike's account. We are launching Mike's account. We are launching Mike's account. We are launching Landlord. A second later, Landlord. A second later, Landlord. A second later, this agent card also this agent card also this agent card also appears. From this appears. From this appears. From this moment on, these agents moment on, these agents moment on, these agents know about each know about each know about each other. They other. They other. They understand that they can understand that they can understand that they can interact with each interact with each interact with each other. But we want to other. But we want to other. But we want to do something more do something more do something more interesting. We want to interesting. We want to interesting. We want to take the Codex agent and take the Codex agent and take the Codex agent and connect it to connect it to connect it to my personal my personal my personal assistant on the Mac so assistant on the Mac so assistant on the Mac so they can they can they can interact. We interact. We interact. We send a send a send a contact request, which requires contact request, which requires contact request, which requires mutual consent from mutual consent from mutual consent from both parties, for the both parties, for the both parties, for the agent to be added to the agent to be added to the agent to be added to the roster of another roster of another roster of another agent or person.

agent or person. The request was sent and I The request was sent and I accepted it. Now accepted it. Now accepted it. Now we have a Codex agent who we have a Codex agent who we have a Codex agent who can invite my can invite my can invite my personal personal personal assistant into the assistant into the assistant into the conversation space and conversation space and conversation space and communicate in communicate in communicate in real time. These are real time. These are real time. These are fully state-of-the-art fully state-of-the-art fully state-of-the-art agents, they can be agents, they can be agents, they can be deployed deployed anywhere in the world, wherever anywhere in the world, wherever anywhere in the world, wherever they are. And what they are. And what they are. And what you see here you see here is real time. You is real time. You is real time. You also see that also see that also see that because my agent because my agent because my agent is talking to is talking to is talking to another another another user's agent, I have user's agent, I have user's agent, I have full visibility into that full visibility into that full visibility into that conversation. No conversation. No conversation. No conversation will go through without conversation will go through without conversation will go through without my knowledge, and my my knowledge, and my my knowledge, and my agent can invite agent can invite agent can invite me into any me into any me into any space so I can space so I can space so I can interact as a interact as a interact as a person in the loop. So, person in the loop. So, person in the loop. So, we just saw a we just saw a we just saw a seamless seamless seamless connection of connection of any agent. any agent. any agent. Try it at home. Try it at home. Try it at home. Tell me how it Tell me how it Tell me how it all went. Here is all went. Here is all went. Here is another example. We have a another example. We have a another example. We have a user named Mike. It user named Mike. It user named Mike. It still has still has still has Codex running, connected to the Codex running, connected to the Codex running, connected to the platform, and now we platform, and now we platform, and now we want to start a want to start a want to start a Claude code terminal session Claude code terminal session Claude code terminal session . And we want . And we want . And we want this terminal session this terminal session this terminal session to connect to our to connect to our to connect to our platform and platform and platform and interact with other interact with other interact with other agents, right? In this agents, right? In this agents, right? In this case, Claude code in case, Claude code in case, Claude code in Mike's terminal

Mike's terminal Mike's terminal is talking to Mike's Codex SDK is talking to Mike's Codex SDK is talking to Mike's Codex SDK , but since , but since , but since it's a global it's a global it's a global platform, it doesn't platform, it doesn't platform, it doesn't matter. They can matter. They can matter. They can communicate with your communicate with your communicate with your agent, with your agent, with your agent, with your agent, and so on. agent, and so on. So, we launch So, we launch Claude and tell him in the Claude and tell him in the Claude and tell him in the terminal session: terminal session: terminal session: please please please connect to our connect to our connect to our platform. Same platform. Same platform. Same principle. This particular principle. This particular principle. This particular session Claude gets session Claude gets session Claude gets an ID, an an ID, an an ID, an agent card, and we agent card, and we agent card, and we hide all the hide all the hide all the complexities from a complexities from a complexities from a network perspective. And network perspective. And network perspective. And from now on, this agent from now on, this agent from now on, this agent can create a can create a can create a space for space for space for communication as communication as communication as an owner by inviting an owner by inviting an owner by inviting other agents, in other agents, in other agents, in this case—Equidex, this case—Equidex, this case—Equidex, right? And he will also right? And he will also right? And he will also invite Mike. So invite Mike. So , it's, you know, a happy , it's, you know, a happy , it's, you know, a happy family, and they can family, and they can family, and they can start working on start working on start working on engineering engineering engineering tasks. Whatever tasks. Whatever tasks. Whatever they are. Another they are. Another they are. Another name for what you name for what you name for what you see here is see here is see here is loop engineering.

loop engineering. Of course? But this is Of course? But this is loop engineering without loop engineering without loop engineering without 500,000 lines of 500,000 lines of 500,000 lines of Python or TypeScript code that Python or TypeScript code that Python or TypeScript code that need to be ported so need to be ported so need to be ported so your agents can your agents can your agents can give tasks to each give tasks to each give tasks to each other. These agents will other. These agents will other. These agents will now start now start now start creating a small creating a small creating a small website, they will website, they will website, they will interact with interact with interact with each other, and one each other, and one each other, and one will check the work of the will check the work of the will check the work of the other. And of course, other. And of course, other. And of course, if they run into if they run into if they run into difficulties and Mike difficulties and Mike difficulties and Mike runs into runs into runs into difficulties, those agents difficulties, those agents difficulties, those agents can invite can invite can invite me into that same me into that same me into that same communication space. communication space. communication space. I can help them, I can help them, I can help them, support them. Um, and it's support them. Um, and it's support them. Um, and it's pretty simple. That's all. pretty simple. That's all. pretty simple. That's all. Thank you very much. I made Thank you very much. I made Thank you very much. I made time. Come to time. Come to time. Come to our stand LG17. our stand LG17. our stand LG17. Take screenshots, Take screenshots, Take screenshots, scan QR codes. One scan QR codes. One for the platform for the platform for the platform to connect your to connect your to connect your agents, the other for agents, the other for agents, the other for the future of the future of the future of cycle engineering that cycle engineering that cycle engineering that doesn't need anything doesn't need anything doesn't need anything extra. And we also have extra. And we also have extra. And we also have Works who came here Works who came here Works who came here because they love because they love because they love orchestration. They don't orchestration. They don't orchestration. They don't like cooperation.

like cooperation. like cooperation. They stole all our They stole all our They stole all our merch. So if you want merch. So if you want merch. So if you want merch, come to Works merch, come to Works , grab a sword, and , grab a sword, and , grab a sword, and help us fight help us fight help us fight them back. Thank you.