500 Skills, Zero Fine-Tuning: LinkedIn's Playbook for AI Agents — Ajay Prakash, LinkedIn

Hey everyone, good morning. Uh Hey everyone, good morning. Uh thanks for being here. Uh I see people thanks for being here. Uh I see people thanks for being here. Uh I see people are still coming. Uh but yeah, my name are still coming. Uh but yeah, my name are still coming. Uh but yeah, my name is uh AJ and uh I am a software engineer is uh AJ and uh I am a software engineer is uh AJ and uh I am a software engineer at uh LinkedIn. at uh LinkedIn. at uh LinkedIn. Today I'm going to be talking about how Today I'm going to be talking about how Today I'm going to be talking about how we are doing context engineering to we are doing context engineering to we are doing context engineering to improve the performance of coding agents improve the performance of coding agents improve the performance of coding agents at LinkedIn. Okay. Uh imagine you are a software Okay. Uh imagine you are a software engineer in a big tech company and your engineer in a big tech company and your engineer in a big tech company and your products are being used by millions of products are being used by millions of products are being used by millions of users on a daily basis users on a daily basis users on a daily basis and you are happen to be on a team which and you are happen to be on a team which and you are happen to be on a team which owns set of very critical services and owns set of very critical services and owns set of very critical services and you are on call right and you get an you are on call right and you get an you are on call right and you get an alert saying that there is an error alert saying that there is an error alert saying that there is an error spike in one of your services spike in one of your services spike in one of your services and uh by the time you're trying to and uh by the time you're trying to and uh by the time you're trying to figure out how to deal with this issue, figure out how to deal with this issue, figure out how to deal with this issue, you take the link to the alert, give it you take the link to the alert, give it you take the link to the alert, give it to a coding agent like Cloud Code or to a coding agent like Cloud Code or to a coding agent like Cloud Code or GitHub Copilot.

GitHub Copilot. GitHub Copilot. While you're trying to figure out um how While you're trying to figure out um how While you're trying to figure out um how to deal with the issue, uh the coding to deal with the issue, uh the coding to deal with the issue, uh the coding agent is working in the background, it agent is working in the background, it agent is working in the background, it will fetch the instructions on how to will fetch the instructions on how to will fetch the instructions on how to debug such issues in your company and debug such issues in your company and debug such issues in your company and identifies that based on that identifies that based on that identifies that based on that instruction, it identifies that this instruction, it identifies that this instruction, it identifies that this alert is happening in uh specific alert is happening in uh specific alert is happening in uh specific service. Then it fetches instruction and service. Then it fetches instruction and service. Then it fetches instruction and context on how to debug that particular context on how to debug that particular context on how to debug that particular service and it will fetch the based on service and it will fetch the based on service and it will fetch the based on those instructions it will take actions those instructions it will take actions those instructions it will take actions like fetching logs metrics and then it like fetching logs metrics and then it like fetching logs metrics and then it uses those logs to identify the root uses those logs to identify the root uses those logs to identify the root cause of the issue right so it cause of the issue right so it cause of the issue right so it identifies based on the error logs where identifies based on the error logs where identifies based on the error logs where the issue is happening and it doesn't the issue is happening and it doesn't the issue is happening and it doesn't just find the root cause it also figures just find the root cause it also figures just find the root cause it also figures out the steps to mitigate the issue and out the steps to mitigate the issue and out the steps to mitigate the issue and once it finds all the det details. It once it finds all the det details. It once it finds all the det details. It summarizes and give it to you saying summarizes and give it to you saying summarizes and give it to you saying this is this is the error and this is this is this is the error and this is this is this is the error and this is the issue and this is the actions that the issue and this is the actions that the issue and this is the actions that you need to take to mitigate. And once you need to take to mitigate. And once you need to take to mitigate. And once you confirm, it also goes ahead and you confirm, it also goes ahead and you confirm, it also goes ahead and takes those actions on your behalf to takes those actions on your behalf to takes those actions on your behalf to mitigate the issue.

mitigate the issue. mitigate the issue. And it doesn't just stop there. Uh it And it doesn't just stop there. Uh it And it doesn't just stop there. Uh it updates your incident management system updates your incident management system updates your incident management system with all the details, error metrics and with all the details, error metrics and with all the details, error metrics and dashboards etc. and also it checks out dashboards etc. and also it checks out dashboards etc. and also it checks out the code and creates a PR for you uh to the code and creates a PR for you uh to the code and creates a PR for you uh to fix the root cause of the issue. All of fix the root cause of the issue. All of fix the root cause of the issue. All of this happens in matter of few minutes this happens in matter of few minutes this happens in matter of few minutes which would have easily taken few hours which would have easily taken few hours which would have easily taken few hours if you were to do it manually. if you were to do it manually. if you were to do it manually. This is not fiction. So this is how This is not fiction. So this is how This is not fiction. So this is how teams at LinkedIn are using coding teams at LinkedIn are using coding teams at LinkedIn are using coding agents as effective co-workers with deep agents as effective co-workers with deep agents as effective co-workers with deep understanding of LinkedIn's internal understanding of LinkedIn's internal understanding of LinkedIn's internal systems and code to help the teams be systems and code to help the teams be systems and code to help the teams be really productive. And this is possible because of a system And this is possible because of a system that we built called as contextual agent that we built called as contextual agent that we built called as contextual agent playbooks and tools at LinkedIn. And playbooks and tools at LinkedIn. And playbooks and tools at LinkedIn. And today I'm going to talk about why we today I'm going to talk about why we today I'm going to talk about why we built the system, how we built it and built the system, how we built it and built the system, how we built it and what are our learnings from the success. To understand why we built the this To understand why we built the this system, we have to go back to the early system, we have to go back to the early system, we have to go back to the early days of coding agents, right? So just days of coding agents, right? So just days of coding agents, right? So just like any other company even at LinkedIn like any other company even at LinkedIn like any other company even at LinkedIn we wanted to use the coding agents to be we wanted to use the coding agents to be we wanted to use the coding agents to be for our engineers and everyone to be for our engineers and everyone to be for our engineers and everyone to be really productive with the AI.

really productive with the AI. really productive with the AI. So we started using started giving this So we started using started giving this So we started using started giving this coding agents to all of the engineers coding agents to all of the engineers coding agents to all of the engineers and the problem was the coding agents and the problem was the coding agents and the problem was the coding agents doesn't really or the w coding doesn't doesn't really or the w coding doesn't doesn't really or the w coding doesn't really work in a large enterprise at really work in a large enterprise at really work in a large enterprise at link um like LinkedIn. link um like LinkedIn. link um like LinkedIn. So the biggest problem is the coding So the biggest problem is the coding So the biggest problem is the coding agent or the LLMs are trained on agent or the LLMs are trained on agent or the LLMs are trained on open-source repos right. So they don't open-source repos right. So they don't open-source repos right. So they don't have the context of how we our uh mature have the context of how we our uh mature have the context of how we our uh mature code bases at LinkedIn or our internal code bases at LinkedIn or our internal code bases at LinkedIn or our internal frameworks or internal systems. frameworks or internal systems. frameworks or internal systems. So what used to happen was the engineers So what used to happen was the engineers So what used to happen was the engineers used to do wipe code or try the agentic used to do wipe code or try the agentic used to do wipe code or try the agentic coding but because it the agents lacked coding but because it the agents lacked coding but because it the agents lacked context they used to hallucinate and context they used to hallucinate and context they used to hallucinate and like get stuck in between or uh even like get stuck in between or uh even like get stuck in between or uh even more dangerous they used to make up more dangerous they used to make up more dangerous they used to make up things which is not correct. So the things which is not correct. So the things which is not correct. So the engineers had to prompt these agents engineers had to prompt these agents engineers had to prompt these agents manually um to do the right thing which manually um to do the right thing which manually um to do the right thing which used to take more time than the manual used to take more time than the manual used to take more time than the manual coding itself. So a lot of people a lot coding itself. So a lot of people a lot coding itself. So a lot of people a lot of engineers went back to manual coding.

of engineers went back to manual coding. of engineers went back to manual coding. So coding agents was not effective So coding agents was not effective So coding agents was not effective to understand the problem uh to get more to understand the problem uh to get more to understand the problem uh to get more perspective. So if you look at the perspective. So if you look at the perspective. So if you look at the LinkedIn stack, we have over thousand LinkedIn stack, we have over thousand LinkedIn stack, we have over thousand repos which make up thousands of uh repos which make up thousands of uh repos which make up thousands of uh microservices and apps microservices and apps microservices and apps and we have a lot of all of these apps and we have a lot of all of these apps and we have a lot of all of these apps and services are built on lot of and services are built on lot of and services are built on lot of internal frameworks and libraries internal frameworks and libraries internal frameworks and libraries and we also have a lot of custombuilt and we also have a lot of custombuilt and we also have a lot of custombuilt infra. For example, we have our own infra. For example, we have our own infra. For example, we have our own databases. We have our own databases. We have our own databases. We have our own experimentation and tracking platform. experimentation and tracking platform. experimentation and tracking platform. We have our own configuration management We have our own configuration management We have our own configuration management system which is purely internal to system which is purely internal to system which is purely internal to LinkedIn and coding agents doesn't have LinkedIn and coding agents doesn't have LinkedIn and coding agents doesn't have any idea about them any idea about them any idea about them and engineers go through a week-long and engineers go through a week-long and engineers go through a week-long boot camp whenever a new engineer joins. boot camp whenever a new engineer joins. boot camp whenever a new engineer joins. So uh just to get familiar with these So uh just to get familiar with these So uh just to get familiar with these systems. So So we looked at this problem and we asked we looked at this problem and we asked we looked at this problem and we asked ourselves the question how can we make ourselves the question how can we make ourselves the question how can we make any coding agent like cursor or cloud any coding agent like cursor or cloud any coding agent like cursor or cloud code or uh GitHub copilot understand our code or uh GitHub copilot understand our code or uh GitHub copilot understand our LinkedIn's internal system so well that LinkedIn's internal system so well that LinkedIn's internal system so well that they can ship the code that our they can ship the code that our they can ship the code that our engineers can trust.

engineers can trust. engineers can trust. By trust I mean the code should be By trust I mean the code should be By trust I mean the code should be correct and also the quality of the code correct and also the quality of the code correct and also the quality of the code should be as good as uh it is written by should be as good as uh it is written by should be as good as uh it is written by an actual engineer. So that is the bar an actual engineer. So that is the bar an actual engineer. So that is the bar we set out and wanted to figure out how we set out and wanted to figure out how we set out and wanted to figure out how do we get there. do we get there. do we get there. So in early 2025 uh last year so uh So in early 2025 uh last year so uh So in early 2025 uh last year so uh Anthropic released MCB and it it quickly Anthropic released MCB and it it quickly Anthropic released MCB and it it quickly became the standard for industry became the standard for industry became the standard for industry standard for building tools to the standard for building tools to the standard for building tools to the agents. We leveraged that and pretty agents. We leveraged that and pretty agents. We leveraged that and pretty early on we built our own internal MCP early on we built our own internal MCP early on we built our own internal MCP and the first tool that we built was uh and the first tool that we built was uh and the first tool that we built was uh code search. code search. code search. So we have a pretty sophisticated code So we have a pretty sophisticated code So we have a pretty sophisticated code search system at LinkedIn where you um search system at LinkedIn where you um search system at LinkedIn where you um engineers can go and search for code. It engineers can go and search for code. It engineers can go and search for code. It will ingest all of uh search for any will ingest all of uh search for any will ingest all of uh search for any code across thousands of repos using code across thousands of repos using code across thousands of repos using keywords and custom filters and rejects keywords and custom filters and rejects keywords and custom filters and rejects etc. So we made that available to the etc. So we made that available to the etc. So we made that available to the coding agents via MCP. This was a really coding agents via MCP. This was a really coding agents via MCP. This was a really powerful unlock because now you don't powerful unlock because now you don't powerful unlock because now you don't have to manually figure out how to do have to manually figure out how to do have to manually figure out how to do better search the agent. You ask a better search the agent. You ask a better search the agent. You ask a question. Hey, how do I set up a question. Hey, how do I set up a question. Hey, how do I set up a particular thing? And the agent can use particular thing? And the agent can use particular thing? And the agent can use the code search tools to figure out the the code search tools to figure out the the code search tools to figure out the right examples of how we do things at right examples of how we do things at right examples of how we do things at LinkedIn and use that to give you answer LinkedIn and use that to give you answer LinkedIn and use that to give you answer and also uh implement it based on the and also uh implement it based on the and also uh implement it based on the its findings. This was really powerful.

its findings. This was really powerful. its findings. This was really powerful. So we added more tools. We added docs, So we added more tools. We added docs, So we added more tools. We added docs, uh, Jiraa, Slack, um, even connected to uh, Jiraa, Slack, um, even connected to uh, Jiraa, Slack, um, even connected to all of our data platforms and even all of our data platforms and even all of our data platforms and even feature flags. So every core tool that feature flags. So every core tool that feature flags. So every core tool that we added to our internal MCP it created we added to our internal MCP it created we added to our internal MCP it created more value by comp it it's almost like a more value by comp it it's almost like a more value by comp it it's almost like a compounding effect because now a compounding effect because now a compounding effect because now a engineer can bring in the PRDS product engineer can bring in the PRDS product engineer can bring in the PRDS product requirement documents and design docs requirement documents and design docs requirement documents and design docs and also their Jira tasks which has mult and also their Jira tasks which has mult and also their Jira tasks which has mult uh different context and use all this to uh different context and use all this to uh different context and use all this to give to the coding agent to automate give to the coding agent to automate give to the coding agent to automate their um or help with their coding. But there was a problem. So you connect But there was a problem. So you connect all these tools but it's not enough all these tools but it's not enough all these tools but it's not enough right? So even with a slightly complex right? So even with a slightly complex right? So even with a slightly complex workflow the agents used to not do workflow the agents used to not do workflow the agents used to not do really well. The problem for example if really well. The problem for example if really well. The problem for example if you give a context it is uh with the you give a context it is uh with the you give a context it is uh with the tools the agent was able to answer tools the agent was able to answer tools the agent was able to answer questions answer basic questions and questions answer basic questions and questions answer basic questions and find code examples but it cannot do a find code examples but it cannot do a find code examples but it cannot do a like a complete job reliably end to end.

like a complete job reliably end to end. like a complete job reliably end to end. The main problem was to do a specific The main problem was to do a specific The main problem was to do a specific job end to end it needs to have a lot of job end to end it needs to have a lot of job end to end it needs to have a lot of tribal knowledge right. So all of for tribal knowledge right. So all of for tribal knowledge right. So all of for example how to fix a particular error or example how to fix a particular error or example how to fix a particular error or how to uh configure how do you debug a how to uh configure how do you debug a how to uh configure how do you debug a particular um error log right so all of particular um error log right so all of particular um error log right so all of this knowledge even though you have this knowledge even though you have this knowledge even though you have access to the tools it is scattered access to the tools it is scattered access to the tools it is scattered across lot of different surfaces for across lot of different surfaces for across lot of different surfaces for example docs wikis and slack example docs wikis and slack example docs wikis and slack conversations conversations conversations um etc and most of the times you may um etc and most of the times you may um etc and most of the times you may have experienced the docs and wikis have experienced the docs and wikis have experienced the docs and wikis might be outdated written and there might be outdated written and there might be outdated written and there might be like duplicate get docs, right? might be like duplicate get docs, right? might be like duplicate get docs, right? So, the problem is the agents even So, the problem is the agents even So, the problem is the agents even though they have access to the tools, though they have access to the tools, though they have access to the tools, they used to get lost. they used to get lost. they used to get lost. The second problem was context overload. The second problem was context overload. The second problem was context overload. As agents use more and more tools, As agents use more and more tools, As agents use more and more tools, their context gets overloaded which their context gets overloaded which their context gets overloaded which means every tool uh tool output it takes means every tool uh tool output it takes means every tool uh tool output it takes up space in the context which will up space in the context which will up space in the context which will eventually cause the agent to compact eventually cause the agent to compact eventually cause the agent to compact its um while it is working compacts it its um while it is working compacts it its um while it is working compacts it its context which causes it to lose some its context which causes it to lose some its context which causes it to lose some of the information. Then it has to do of the information. Then it has to do of the information. Then it has to do all over again.

all over again. all over again. And the third problem was even though And the third problem was even though And the third problem was even though even if the agent was able to figure out even if the agent was able to figure out even if the agent was able to figure out all these uh details it can it doesn't all these uh details it can it doesn't all these uh details it can it doesn't have a way to retain this information it have a way to retain this information it have a way to retain this information it doesn't have a a durable memory. So doesn't have a a durable memory. So doesn't have a a durable memory. So every time a engineer asks the uh agent every time a engineer asks the uh agent every time a engineer asks the uh agent to do a certain task they have to start to do a certain task they have to start to do a certain task they have to start from scratch. So how do we solve this problem? So we So how do we solve this problem? So we give this instructions right away right? give this instructions right away right? give this instructions right away right? So we built a system, we invented a So we built a system, we invented a So we built a system, we invented a system in early 2025 called as uh system in early 2025 called as uh system in early 2025 called as uh playbooks where we not only provide the playbooks where we not only provide the playbooks where we not only provide the tools to the agents via MCP, we also tools to the agents via MCP, we also tools to the agents via MCP, we also allow the agents to access these allow the agents to access these allow the agents to access these instructions and prompts via MCP. We instructions and prompts via MCP. We instructions and prompts via MCP. We call it playbooks. And playbook call it playbooks. And playbook call it playbooks. And playbook it just appears just like any other it just appears just like any other it just appears just like any other regular tool. regular tool. regular tool. uh and they have uh names and uh and they have uh names and uh and they have uh names and description on what it does and the description on what it does and the description on what it does and the agent can decide to invoke that playbook agent can decide to invoke that playbook agent can decide to invoke that playbook just like any other regular tool and just like any other regular tool and just like any other regular tool and when the uh playbook is invoked the when the uh playbook is invoked the when the uh playbook is invoked the instructions and the context within that instructions and the context within that instructions and the context within that playbook are returned as the tool output playbook are returned as the tool output playbook are returned as the tool output to the coding agent.

to the coding agent. to the coding agent. So that way the agents have both tools So that way the agents have both tools So that way the agents have both tools as well as instructions on how to use as well as instructions on how to use as well as instructions on how to use tools to set up a or uh uh perform a tools to set up a or uh uh perform a tools to set up a or uh uh perform a task. Right? For example, uh if the uh task. Right? For example, uh if the uh task. Right? For example, uh if the uh engineer goes and ask like how do I set engineer goes and ask like how do I set engineer goes and ask like how do I set up a airflow DAG uh at LinkedIn? The up a airflow DAG uh at LinkedIn? The up a airflow DAG uh at LinkedIn? The agent will first decide okay so I have a agent will first decide okay so I have a agent will first decide okay so I have a playbook for creating um that specific playbook for creating um that specific playbook for creating um that specific task and it will use that first fet uh task and it will use that first fet uh task and it will use that first fet uh uses that playbook to get the uses that playbook to get the uses that playbook to get the information and then it calls the information and then it calls the information and then it calls the necessary uh follows that instructions necessary uh follows that instructions necessary uh follows that instructions and calls the relevant tools to get the and calls the relevant tools to get the and calls the relevant tools to get the job done. This was really powerful job done. This was really powerful job done. This was really powerful um mainly because now anyone at LinkedIn um mainly because now anyone at LinkedIn um mainly because now anyone at LinkedIn can go ahead and create a set up a can go ahead and create a set up a can go ahead and create a set up a playbook and check it into a repository playbook and check it into a repository playbook and check it into a repository and make it available for everyone else and make it available for everyone else and make it available for everyone else at LinkedIn. So as people started creating more So as people started creating more playbooks so we wanted so this is um one playbooks so we wanted so this is um one playbooks so we wanted so this is um one of two foundational principles we want of two foundational principles we want of two foundational principles we want everyone to follow when creating a everyone to follow when creating a everyone to follow when creating a playbook. The first one is a playbook playbook. The first one is a playbook playbook. The first one is a playbook should be self-contained which means it should be self-contained which means it should be self-contained which means it should do a very specific task only um should do a very specific task only um should do a very specific task only um for example um if it is for setting up a for example um if it is for setting up a for example um if it is for setting up a airflow DAG it should be about the airflow DAG it should be about the airflow DAG it should be about the instruction and the construct should be instruction and the construct should be instruction and the construct should be about uh one specific task. This helps about uh one specific task. This helps about uh one specific task. This helps the agents pick the right playbook for

the agents pick the right playbook for the agents pick the right playbook for the right task. And the second uh most the right task. And the second uh most the right task. And the second uh most important one is to break a big playbook important one is to break a big playbook important one is to break a big playbook into multiple smaller playbooks. into multiple smaller playbooks. into multiple smaller playbooks. So this has um and reference those So this has um and reference those So this has um and reference those smaller playbooks from a bigger smaller playbooks from a bigger smaller playbooks from a bigger playbook. This is a really powerful playbook. This is a really powerful playbook. This is a really powerful um principle because just like um so it um principle because just like um so it um principle because just like um so it has two main advantages, right? So the has two main advantages, right? So the has two main advantages, right? So the first one is uh reusability. So if you first one is uh reusability. So if you first one is uh reusability. So if you have a small self-contained playbooks, have a small self-contained playbooks, have a small self-contained playbooks, it can be used from multiple uh it can be used from multiple uh it can be used from multiple uh reference from multiple playbooks and if reference from multiple playbooks and if reference from multiple playbooks and if you um the another big advantage is you um the another big advantage is you um the another big advantage is progressive discovery of um context progressive discovery of um context progressive discovery of um context which means the agent only when it needs which means the agent only when it needs which means the agent only when it needs to read a smaller playbook instead of to read a smaller playbook instead of to read a smaller playbook instead of reading the entire all of the playbooks reading the entire all of the playbooks reading the entire all of the playbooks at once it can uh progressively go and at once it can uh progressively go and at once it can uh progressively go and read the playbooks as it wants. So this read the playbooks as it wants. So this read the playbooks as it wants. So this is the same concept as skills as well. is the same concept as skills as well. is the same concept as skills as well. So playbooks are very similar to uh So playbooks are very similar to uh So playbooks are very similar to uh skills but we developed this entire skills but we developed this entire skills but we developed this entire system around playbooks even before system around playbooks even before system around playbooks even before skills was a thing. And uh playbooks are skills was a thing. And uh playbooks are skills was a thing. And uh playbooks are a little bit more nuanced because it a little bit more nuanced because it a little bit more nuanced because it helps us um it helps us seamlessly helps us um it helps us seamlessly helps us um it helps us seamlessly capture all of the organizational capture all of the organizational capture all of the organizational context and service via MCP without much context and service via MCP without much context and service via MCP without much of a setup.

And another cool thing about this And another cool thing about this playbooks is this self-improving loop. playbooks is this self-improving loop. playbooks is this self-improving loop. So u you have uh engineers creating So u you have uh engineers creating So u you have uh engineers creating these playbooks and checking into the these playbooks and checking into the these playbooks and checking into the repository. And one of the main problem repository. And one of the main problem repository. And one of the main problem with any knowledge base is it gets with any knowledge base is it gets with any knowledge base is it gets outdated. How do you uh the biggest outdated. How do you uh the biggest outdated. How do you uh the biggest problem is how do you keep the context problem is how do you keep the context problem is how do you keep the context fresh right? So great thing about agents fresh right? So great thing about agents fresh right? So great thing about agents is they can improvise. is they can improvise. is they can improvise. So we have we encourage the agents to So we have we encourage the agents to So we have we encourage the agents to whenever they use a particular playbook whenever they use a particular playbook whenever they use a particular playbook at the end of the session to identify at the end of the session to identify at the end of the session to identify the learnings. So any outdated the learnings. So any outdated the learnings. So any outdated information or any discrepancy or any information or any discrepancy or any information or any discrepancy or any missing information and we also missing information and we also missing information and we also encourage the agents to figure out how encourage the agents to figure out how encourage the agents to figure out how to improve the playbook and use that to improve the playbook and use that to improve the playbook and use that context to check it uh to update the context to check it uh to update the context to check it uh to update the playbooks check out the uh repository playbooks check out the uh repository playbooks check out the uh repository and update the playbooks and create a PR and update the playbooks and create a PR and update the playbooks and create a PR and that once it gets upload it gets uh and that once it gets upload it gets uh and that once it gets upload it gets uh the playbooks gets updated right this the playbooks gets updated right this the playbooks gets updated right this creates a really seamless flywheel of a creates a really seamless flywheel of a creates a really seamless flywheel of a self-arning loop.

So what does the architecture of a MCP So what does the architecture of a MCP server looks like? server looks like? server looks like? So this particular system we have one So this particular system we have one So this particular system we have one local MCP server and it is automatically local MCP server and it is automatically local MCP server and it is automatically installed on all of the LinkedIn laptops installed on all of the LinkedIn laptops installed on all of the LinkedIn laptops by default. So if you join LinkedIn and by default. So if you join LinkedIn and by default. So if you join LinkedIn and you get a laptop, it is pre-installed you get a laptop, it is pre-installed you get a laptop, it is pre-installed and any updates to the uh MCP server or and any updates to the uh MCP server or and any updates to the uh MCP server or the playbooks or the tools, it the playbooks or the tools, it the playbooks or the tools, it automatically gets updated every 1 hour automatically gets updated every 1 hour automatically gets updated every 1 hour on all the laptops. on all the laptops. on all the laptops. And we have a concept of two local And we have a concept of two local And we have a concept of two local playbooks and central playbooks which playbooks and central playbooks which playbooks and central playbooks which means so central playbooks are the means so central playbooks are the means so central playbooks are the playbooks which are crosscutting um in playbooks which are crosscutting um in playbooks which are crosscutting um in nature, right? So you have um these nature, right? So you have um these nature, right? So you have um these playbooks apply for multiple uh playbooks apply for multiple uh playbooks apply for multiple uh repositories not just one uh code repositories not just one uh code repositories not just one uh code repository and then you have local repository and then you have local repository and then you have local playbooks where it these are the playbooks where it these are the playbooks where it these are the playbooks which are very specific to playbooks which are very specific to playbooks which are very specific to your code repository and you can just your code repository and you can just your code repository and you can just have them checked in with your repo um have them checked in with your repo um have them checked in with your repo um uh with your repo and when only when the uh with your repo and when only when the uh with your repo and when only when the coding agents are working in your repo coding agents are working in your repo coding agents are working in your repo those playbooks will be automatically those playbooks will be automatically those playbooks will be automatically picked up. So this helps us scale the picked up. So this helps us scale the picked up. So this helps us scale the local playbooks which are very specific local playbooks which are very specific local playbooks which are very specific to repo without having to worry about um to repo without having to worry about um to repo without having to worry about um uh changing the uh central repository uh changing the uh central repository uh changing the uh central repository and also this is one MCP server which is and also this is one MCP server which is and also this is one MCP server which is serving all of the uh playbooks and serving all of the uh playbooks and serving all of the uh playbooks and tools. So this man helps us do a lot of tools. So this man helps us do a lot of tools. So this man helps us do a lot of uh central things like seamless uh central things like seamless uh central things like seamless authentication, telemetry and u that we

authentication, telemetry and u that we authentication, telemetry and u that we can use for learning to make the whole can use for learning to make the whole can use for learning to make the whole uh ecosystem better. uh ecosystem better. uh ecosystem better. You may be wondering like how many tools You may be wondering like how many tools You may be wondering like how many tools and playbooks it can support, right? Uh and playbooks it can support, right? Uh and playbooks it can support, right? Uh so this is a common problem with MCP. we so this is a common problem with MCP. we so this is a common problem with MCP. we cannot scale it beyond 30 or 40 tools cannot scale it beyond 30 or 40 tools cannot scale it beyond 30 or 40 tools without degrading the uh context or without degrading the uh context or without degrading the uh context or degrading the performance of the system. degrading the performance of the system. degrading the performance of the system. So what we do is instead of uh surfacing So what we do is instead of uh surfacing So what we do is instead of uh surfacing all of these playbooks and tools through all of these playbooks and tools through all of these playbooks and tools through MCP we replace them with three meta MCP we replace them with three meta MCP we replace them with three meta tools. So the first one is search. The tools. So the first one is search. The tools. So the first one is search. The agent first uses this tool to search for agent first uses this tool to search for agent first uses this tool to search for the relevant tools and playbooks using the relevant tools and playbooks using the relevant tools and playbooks using keywords and tags. U so we also control keywords and tags. U so we also control keywords and tags. U so we also control the system instructions. Um so every the system instructions. Um so every the system instructions. Um so every coding agent is preconfigured with coding agent is preconfigured with coding agent is preconfigured with system instruction on how to use these system instruction on how to use these system instruction on how to use these tools and how to use the search really tools and how to use the search really tools and how to use the search really efficiently. And once it finds the uh efficiently. And once it finds the uh efficiently. And once it finds the uh right set of tool or playbook, it can right set of tool or playbook, it can right set of tool or playbook, it can then get the more details about that then get the more details about that then get the more details about that particular tool using get schema and particular tool using get schema and particular tool using get schema and then execute that tool or uh playbook.

then execute that tool or uh playbook. then execute that tool or uh playbook. So this has allowed us to scale uh to So this has allowed us to scale uh to So this has allowed us to scale uh to thousands of tools in playbook. Um so this is u the growth chart. So now Um so this is u the growth chart. So now we have over 8,000 users daily um using we have over 8,000 users daily um using we have over 8,000 users daily um using the system daily. using uh tools and the system daily. using uh tools and the system daily. using uh tools and playbooks. So we have over,300 playbooks. So we have over,300 playbooks. So we have over,300 uh tools and over 600 uh playbooks uh tools and over 600 uh playbooks uh tools and over 600 uh playbooks and it's not not just engineering right. and it's not not just engineering right. and it's not not just engineering right. So it is uh not just engineers but also So it is uh not just engineers but also So it is uh not just engineers but also product managers, designers, uh TPMS. So product managers, designers, uh TPMS. So product managers, designers, uh TPMS. So across different functions they are across different functions they are across different functions they are using the tools and bringing their using the tools and bringing their using the tools and bringing their playbooks uh to automate their uh playbooks uh to automate their uh playbooks uh to automate their uh workflows. uh so I'll leave you with this takeaway uh so I'll leave you with this takeaway uh key takeaways that can based on our uh key takeaways that can based on our uh key takeaways that can based on our learning the first one is the system was learning the first one is the system was learning the first one is the system was successful because we thought about successful because we thought about successful because we thought about quality and reliability uh from day one quality and reliability uh from day one quality and reliability uh from day one right so even uh before creating a MCP right so even uh before creating a MCP right so even uh before creating a MCP server uh we thought okay the our server uh we thought okay the our server uh we thought okay the our fundamental principle should be how do fundamental principle should be how do fundamental principle should be how do we ensure not just productivity but how we ensure not just productivity but how we ensure not just productivity but how do we ensure the quality and also do we ensure the quality and also do we ensure the quality and also reliability of the system so that it reliability of the system so that it reliability of the system so that it doesn't degrade um as we move fast.

doesn't degrade um as we move fast. doesn't degrade um as we move fast. Uh and the second one was u the build Uh and the second one was u the build Uh and the second one was u the build the right infrastructure for agents. Uh the right infrastructure for agents. Uh the right infrastructure for agents. Uh in a large enterprise like LinkedIn, in a large enterprise like LinkedIn, in a large enterprise like LinkedIn, it's not enough just enough to give all it's not enough just enough to give all it's not enough just enough to give all of the engineers the all the latest and of the engineers the all the latest and of the engineers the all the latest and greatest tools and models. Uh these are greatest tools and models. Uh these are greatest tools and models. Uh these are they are not very effective if you don't they are not very effective if you don't they are not very effective if you don't build the right infrastructure for the build the right infrastructure for the build the right infrastructure for the agents to operate within your agents to operate within your agents to operate within your enterprise. enterprise. enterprise. Um yeah that's my time. Thank you for Um yeah that's my time. Thank you for Um yeah that's my time. Thank you for attending and feel free to connect with attending and feel free to connect with attending and feel free to connect with me on [applause] LinkedIn.