从实验走向生产：Credit Karma 如何治理企业智能体

Brief Description

Paving the Road to Production: Governed Enterprise AI Agents — Harish Gaggar， Credit Karma

Table of Contents

• From experiments to enterprise agents
• Missing production foundations
• The agent and supporting frameworks
• Registration, local development and the paved path
• Review, containers and Kubernetes deployment
• Reviewed custom MCP servers
• MCP gateway authentication and routing
• Logs, traces and gateway metrics
• LangGraph policies and human approval
• Threads, runs and memory checkpoints
• Agent tracking and evaluation records
• Filtering sensitive inputs
• Context optimization and token reduction demos
• Build the framework around every agent

From experiments to enterprise agents

Hello everyone. Um, I would like to walk you through that how we can transition in AI agents from an experiments to the real enterprise environment. So, before I begin, a quick introduction about me. My name is Sharish Gokar. I'm a senior staff software engineer at Intuit Credit Karma where I built and own the AI analytics infrastructure and services. Um, and beside that, um, I'm advisor, uh, author, researchers, um, and also a program committee members for various organizations where I mentor the young researchers as well. All right. So, let's, uh, walk through the, um, agenda here. So, first we would like to talk about what are the challenges we have when we build an AI agents and deploy into the production systems and how we, uh, can build a different framework to address that, uh, different challenges here.

Um, then how the architecture looks like for it. Um, then, um, how you can build your own MCP servers, um, and MCP gateway. And then some monitoring on top of that gateway as well. Um, then further we're going to talk about what are the different type of agent framework you can have, um, and then, um, some additional framework around managing the memories, tracking and eval, um, how you can block the PII data, uh, going before the LLMs, and then context optimizer to save your token, right?

Missing production foundations

So, without delay, let's, um, let me walk you through the challenge. Um, of course, like there are many agents, uh, uh, generally get failed in productions because their foundation controls are missing. Um, and the key reason behind is that, um, you know, like, oh, most of the people are missing the gateway to make sure that agent can, you know, communicate to the gateway to get authenticated first with the right auth providers so that they can redirect to the right MCP servers, right? Second, that, um, MCP servers are missing the right tool to execute your query. Third, do we have like enough tracking and evaluation system to make sure that every query when you are sending to the agents, whether they've been tracked or not, and are you evaluating the quality response or not, right? Do we have any in memory management system or not in terms of let's say you are doing the interaction with agent for long or for many days, is it able to recall that what you have been asked in the past, right?

And then there are other framework, of course, like do you have a context of the measures to save the token, or do you have inner production in place to make sure that your any secrets should not go into LLM or tracking or tracing or any memory DB or so, right? So, a lot of lot of open questions. So, in order to resolve this, we are not going to, you know, like build a magical agent here in this demo. We're going to talk about like different framework that you can build to address this.

The agent and supporting frameworks

So, first is the agent framework. This is where like we use the like we choose to use the LangGraph framework, which is provided by LangChain community. This is help us in building the task-specific node based on your questions, and which is easy to plug in any like marketplace plugins or skills and prompts to make sure that agents respond efficiently based on your query. Further, the second part is we are also building the MCP gateway to make sure that whenever agent make the call or any service account from that agent make the call, make sure that they are securely get authenticated to that MCP gateway, and then get redirected to the right MCP servers. Third is the customized MCP, which means that you are building your own MCP servers with the right tool in place to execute your query. And then finally some audit on top of what actually operation you're performing, right?

And then we talk about some additional framework around you know like how you can block the PII secrets, how you going to store the tracking eval into the cloud, and how you going to process the memories, especially any threads and checkpoints on every context, and then also how you can

Registration, local development and the paved path

optimize overall your context to make sure that you don't don't burn too many tokens when you run your query. All right? So, we talk about now the agent infrastructure. So, the Firstly, we'll go through the local development. The way the approach should be is that you know like whenever you build an agent, right? First you have to register your agent into some system. You usually tell that hey look, this is my the use case and these are the use case going to use for this particular vertical or system, and you feed all those information and generate a ID for your agent, right? Once you have that, then you can have a dev portal, internal portal where you can ingest all the details saying that hey look, this is my agent use case, this is my agent ID, this is the framework I'm going to use either LangGraph, LangChain or so.

These are my language I'm going to use either Python or TypeScript, and then you can choose you know like your skills or prompts, and then you submit your information, right? To this. Once you submit this, the dev portal is going to create a repository, which is going to contains not just the agent framework like your LangGraph agent framework, but is also going to contains your memory framework, your tracking eval framework, right? and then your context optimizer framework, and your PII blocking framework as well. So, the reason for that is because not necessarily all your developers are taking care of all this framework, right? They just sometimes you know create the framework what generally like AI assistant tools are generally the right? Or they stick to one or two framework, right? So, we just want to make sure that when you roll out such framework, make sure they are consistent across the company um and they can take care of uh and they can be centralized in one place so you can visualize their results as well, right?

So, once you have that repository, so developers basically, you know, clone that repository, they can build their agent as a Docker image and containers and test their agent input and output through the query. And they can also traces on their local memory, maybe cloud span evaluator here. And you can also store this tracking and eval in some of your DB, maybe SQLite or maybe BigQuery um to see to make sure that agents are fully working on your local, right? So,

Review, containers and Kubernetes deployment

once your agents are running in local, um then you're ready to go ahead and generate a PR and then you can once you get your two PR approval, then it's go to the CICD pipeline, right? So, the PR is basically going to firstly is going to uh register your agent into some centralized agent registry system, which provide a length with URL to it to make sure that you see all the traces. And then the CICD pipeline is basically, you know, whatever the local Docker container you built here is going to build that container into production system. Uh it will scan that image, make sure that there's no vulnerability. Um if everything looks great, security approved, you can move that image back into the registry, right? Further, you have your own deployment platform or orchestrator platform which going to deploy that agent into the Kubernetes cluster in a given namespace, right?

So, the good part here is that this dev portal is not just creating the repository, um right? It also do create um bunch of other production resources, for example, the workload identity, um right? And then memory DB or tracking and eval tables and then agent secret management, right? Um so, it based on whatever the name you provide, it not just again it like not just repository, it all these different resources as well. So, what happened is when this internal orchestrator when it deploy the agents, the agent already know that hey, what workload identity I supposed to work with, right? That binding should already happen. So, once this K to GSA binding happens, the agent will start communicating to all these resources. Once everything in place, um agent is further going to communicate to the rest of your endpoint services like for example, AI gateway or MCP servers or length fuse for tracing or some other API endpoints.

Reviewed custom MCP servers

Right? So, now firstly, we talk about the MCP server as well. So, we talk about architecture. Now, we're going to the different component. So, MCP server is something that you can definitely use something which is outside available, but if you're working on air gap environment, it's highly recommended that you build your own MCP servers. So, um yeah, uh and you can use the uh I will not go with the definition. Everyone knows about it. So, the way the framework has been built here is that uh you you can leverage the fast MCP server package um or wrapper to um you know, build that MCP servers and you can create a different set of tools inside that a fast MCP wrapper to construct your um you know, the overall servers. And then you make sure that, you know, like every tool when you define in MCP server, they had to be fully reviewed by your own internal security team to make sure that only authorized tool can be go in production.

And that's how you can you know, enforce some filters, limits, and block the fields which is really not required to go to LLM and get you the information which is you're not supposed to get, right? Um and third part is like a of course, you need to make sure that you build your MCP server in a way that whenever the traffic is coming from your MCP client to MCP gateway to MCP servers, either it's coming from user service account, they has to be authenticated to the MCP gateway first and so that they can be access your MCP tool to get the right information.

MCP gateway authentication and routing

All right. Now, the gateway part. So, the how the gateway works. So, this is again, you can build it your own internal MCP gateway if you are again working on your air gap environment. So, the way it work is you are working on cloud assist like local, you know, code assistance tool or maybe your agent, right? So, the moment you submit your first query to your agent, it will send that one-time authentication request to the MCP gateway, right? MCP gateway recognized and take your request. First, it will save your session state into the Redis, right? And then it will, you know, return back the JWT access token back to the your client, right? Once client will receive that JWT token, it will know that what MCP server needs to call, right? So, it will submit that tool {slash} call request on a given name space {slash} MCP, meaning let's say Artifactory or GitHub {slash} MCP with the auth bearer JWT token into that.

So, once MCP gateway receive that, it will understand that what name space, meaning MCP server needs to look up for, and then it will unwrap that JWT token, and then it will attach that user token in that request and send it to the respective MCP servers. Now, once MCP server receive it, it will make that respective API calls to that endpoint of that, you know, to the service and then it will exchange that ID token and return the response based on whatever scope that user is authorized to see, right? So, once the user receive the token, you know, MCP server basically, you know, unwrap that all the data, make the payload which is compatible for MCP gateway to receive, and then finally the final payload will send back to the client to see the data points. So, what it shows is that, you know, like we build this MCP gateway to make sure that we have a centralized authentication and routing hub that securely connect any coding assistant tool or agents to the endpoint via this MCP gateway and MCP server.

This There

Logs, traces and gateway metrics

should not be direct access provided um on any endpoints. They has to go through the MCP gateway through proper authentications uh before they're reaching to the MCP server and the endpoints. All right? So, now the part is the gateway monitoring, right? We talk about again, we talk about architecture, we talk about MCP server, MCP gateway. Now, it's a monitoring. Again, the AI agent cannot be trusted without continuous monitoring, evaluation, and feedback, right? This is why we are building the MCP gateway uh monitoring. So, the way the monitoring work is usually like the moment you submit any calls to your agent, uh it will send all the operational log into the log store. You can use ELK, Loki, or CloudWatch also, right? And generally, this contains the information like show me like all last 401s in my last 10 minutes call, right?

And what did OAuth do when the token refresh failed, right? So, those kind of requests can goes to the log store. Now, there are some, you know, traces and metrics data um like real-time data that goes to the OpenTelemetry as well. Um OpenTelemetry again is open source solutions, which going to collect all your traces um and metrics and it's going to process the data and redirect to the back-end system, right? And in this case, the traces will go to the Tempo. This is again an open source solutions, which going to store all your traces in terms of like traces in the sense like, you know, like information around like which tool was slow on this request, whether this call is coming from agent or IDE, uh what was the exact error um on this failed call or so, right? So, these kind of data are going to store into the Tempo.

My bad. Okay. And then the metrics is about the Prometheus is all about the information about like what are P95 latencies for a given tool, right? Whether my auth rejection is spiking on a particular time or not, right? So, those kind of metrics is going to store in Prometheus. So, once you get all these data point like traces and metrics, this entire aggregate data actually save into the graph now for your visualizations.

LangGraph policies and human approval

All right. So, the MCP server gateway and monitoring is covered. Now, we're going to talk about the framework. So, the agent framework, right? So, we use LangGraph for most of the cases and is basically LangGraph again it's a provided by LangChain community. It's provide a graph based architecture that support loops, branching, state management for complex multi-agent AI workflows, right? And the way it work is like the moment you submit your first query, we do interpret you know, like your question first, like what type of questions it is and what response is looking for and based on that it's create a plan, right? Then it will send for classify classify area where it will determine that hey, whether it is just like a read operations or whether it is just like a you know, modification level changes coming up or whether you really want to delete something, right?

Or some sensitive operations. Based on that it adjust the overall structure and then plan accordingly. The next step is plan is actually itself is a sub graph where you do the major of your operations. Meaning that in the plan like once you know what operation you're performing, it goes to the like let's say in a fetch evidence section where it actually call that whatever the question you ask here, after all interpret and classify, um it actually go ahead and make the call to the MCP gateway, uh make sure they get authenticated, um and then it make the call to the MCP server using the right tool, um and get the data, and then generate that evidence, right? Once you validate all your finding, then it will check the policies. Uh it will check whether it is a read-only, then go ahead and execute, and I will save all the record points into our audit logs.

But if it is a right and sensitive operations, definitely it will go through the HITL process where you wait for an approval. If it is approved, you go ahead and execute. If it is get rejected, in either case it's going to save into the audit log. Right? So, if we combine all this together, like MCP gateway, MCP server, um and the agent, right? This is how things looks like. So, ultimately, whatever the agent you're building, whatever the call uh you're submitting, it will go through the MCP gateway for authentications, and then it will redirect to the right MCP servers to execute uh you know, your query on a given tool, uh so that it make the API call to that endpoint and get the response. Right? All right. If it If we think from the infrastructure perspective, this is how it looks like. So, agent is communicating to the MCP gateway, goes to the MCP server, and to the endpoint.

All right? So, now the uh we talked a lot about MCP gateway, MCP servers, um the monitoring, um and the agent framework. Now we talk

Threads, runs and memory checkpoints

about uh three and four additional framework that we should have to take care of when building the end-to-end ecosystem, right? One is them is agent memory management. Again, it's a uh you know, it's a it's refer It's basically a process where you can save and restore the state of your workflow over the time, right? Ensuring that it's provide you a durability and seamless recovery across across your every query run. So, the way it work is like whenever you submit any query to your agent, it actually generate a thread, right? On that given thread, when you ask your first question, it will going to create a run ID for that. So, the moment you ask some deeper questions, it will go on a particular node in that langraph and execute that query. And on that node, it going to save that respective checkpoint here, right?

Now, if you ask a follow-up questions on the same thread, it going to create another run ID. And again, depend on your follow-up question, it go to a specific node and save the checkpoint over there, right? And all this data for your local development, you can use the SQLite SQLite DB, but in production, unless in GCP, you can use Cloud Spanner. And in respective databases, you can use in Azure Azure and AWS as well, right? So, just walk you through the quick demo here on the memory. So, for example, I have, you know, like just build this agent locally and then, you know, trying to run this query on SQLite DB to see if there is any data or not. So, I'm just, you know, initially I don't have any data. Now, I submit my query, you will see that I received the one record, which is saying that, "Hey, this is my thread ID.

This is the question I have been asked." And exact time, you know, when this query was executed. Now, if I run the next query to see that, "Hey, how many messages I've been produced on a given thread?" You can get this information by running the further queries, right? Now, if I would like to run to see that, you know, like all the metadata level information on my query, I can run my query again and And can see the detail description around, you know, like how many tool I get invoked, what is my thread ID, what is my run ID, um you know, like and all those informations. Now, if I want to also see that, hey, what tool have been executed when I run this query, um you can go ahead and see the exact tool name as well on that given MCP servers. Now, further I will also want to see that how the progression was happened on my lang graph framework, that also you can see by um running this type of query on your memory DB.

You can see that here like different kind of, you know, nodes been called when actually, um you know, that query has been run.

Agent tracking and evaluation records

All right? Now, the tracking and eval. So, just to make sure that um there is no confusion here like the we talked about earlier the MCP gateway monitoring, this is the agent level tracking. So, we do this because we really want to track that what query you have been submitting to the agent um and what the response is coming, what alarm model you have been used, what's the cost of your token, um how many token you have been burned, and what's your, you know, evaluation criteria looks like. What is the quality of your response or not, right? So, you need to store that data from the day one. And imagine that if you have a thousand plus agent, every agents will have their own um you know, um agent tracking and eval um project and the tables, right? So, if you combine, let's say if you have 100 agents, you will have 100 different projects and each project has their own respective tracking and eval tables.

Now, tomorrow, 6 months down the line, if somebody ask you that how your agent has been performed, what was the response, what's was the quality, how much token you have been burned, you can aggregate all this data and you can create executive dashboards around it to see um with to see the what is the ROI in building so many agents, right? So, let me walk you through the the tracking and eval um demo here. So, um here like if I look at uh this one, so first like, you know, like if you see that there are six tables are listed here. One is a job tracking table, um and which is contains most of this like information around what query like what is your query, what's the response and everything. And there are a bunch of other tables as well like your quality related and your tool invocations and then node level executions and all that.

So, in this case I'm going to submit my first query on my agent um and you will see that um there will not be uh any data in the first place uh because it takes some time. So, I will refresh this again. And this time I will see that when I refresh, I do see that for a given query I get my request ID, my session ID, my agent name, um you know, what question I've been asked, um and what was the output um of that uh question, uh whether it has completed or not, and then what is the time it took, what is the model I've been used, um and then some of the middle level uh middle level informations, right? Further, if I look for the quality, again there is a eight quality checks per run, right? And you can see that this one row each for, you know, like success, like latency, cost, tool accuracy, safety, etc.

Like all this has been stored here, right? Now, the eval scores. This is again this going to be a one row. This is going to be a scoreboard for your given run, which is basically varies between zero to one, which is scores uh you know, like all the quality dimension in a single row, right? Now, the quality metrics. So, again it's a deep dive on that particular one run. Basically a nested field for the response quality, safety, latency, cost, uh and the tool stat in one white row, right? Now, the tool invocations, so every tool call, which MCP tool ran, how long it took, and whether it succeeded or not. Right? And now the tool agent trajectory, which is talk about step-by-step story of the run, each LLM call, and the graph state in order, like plan, summarize, etc. All right. So, now

Filtering sensitive inputs

uh the next is the agent PIR treatment. This is very, very important because many times it happens that when you're working in agents or uh this thing, um sometimes there's a possibility that there are some secrets being ingested as a part of your query, right? Let's say you're uploading some doc which is has some sensitive informations um and sometimes while typing you may have some API keys or so. So, we have a framework which define that different sort of data types and their severity level. If your severity is above 90% or so, uh we will block and redact that uh secrets uh immediately. So, that uh make sure that that secret is never goes to LLM models, right? Neither it will goes to your memory DB, or nor it will go to tracking and eval tables, or tracing uh systems as well. So, the goal here is to just block this sensitive data as it is the moment it get ingested, right?

Moment moment you type.

Context optimization and token reduction demos

All right. Um the last framework is all about now is the context optimizer. This is one of my favorite area where like um you know, uh whenever you submit, it's very important, right? Whenever you submit too many queries, sometimes you upload a lot of the um data points and you upload bunch of file systems or so, right? It's very important that time is um you know, like you make sure that you save your token in every query. So, we build in context optimizer framework where you actually shape your query and then you prune it before it go to LLM. In the pruning system, uh we do select the payload based on the token limit. Then we re-rank that, "Hey, which is the most relevant relevant information that we should target for when you submit this huge giant, you know, context?" And then we dedupe that, uh remove all the duplicates, and create a very unique set of uh very curated data like uh I mean the main data for the LLM.

And then we finally compress this large payload before it goes to the LLM. So, let's see this uh demo here. Uh this will be quick. So, for example, let's say I will submit my first query, "Hey, find me all the artifacts in my JFrog Artifactory." Um you see that first when before even I not submitted any query yet, there is no data here. So, the moment I submit this first query, and if I refresh my query here, um and then you can see that the record has been added. So, you can see there is a 24,000 token has been used. I will go back a little bit um just to show the thing. So, you can see this 24,000 token has been used for given questions. Now, if I ask uh now I enable the context optimizer behind the scene and I'm rerun the exact same query again. So, this time you will see that uh you know, for the same query, if I refresh this run, and you will see that uh this time it used just 2,400 token.

Right? For exact same query. So, which is almost like 90% saving. I can rerun the same demo on the Slack bot as well where I'm asking Slack bot, "Hey, tell me all my Docker images across all the repositories in Artifactory." When I run this query, there is no data. Now, I'm going to submit my query. So, once this query will run, um and if I run my query again here, and you will see the record get created. It's again used like 24,000 for this this particular query. Now, what I will do is I will turn on my optimizer behind the scene and I rerun the same query in the Slack bot. Um and then if I run this query again, um it will show like it exactly used 1,500 almost token, which is again like 90% saving. In most of the cases, we get like 70 to 78% token saves. So, imagine that you have thousands of um you know, like um um if you have thousands of like agents, which is running like you know, many queries per day, you imagine how much token you can save with this framework.

All right. So, now the

Build the framework around every agent

key takeaway here is uh you know, like um for us is like make sure that we build an enterprise framework, um not just an agent, right? Which make sure that you have all the framework underneath. Don't try to be in rush to building some agent quickly and deploy ship it. Make sure that it has to go through all the framework all together, so that you should not realize a letter state that, you know, I should have this framework in place. Right? In in the very first place. Second is like your Make sure that you have a proper gateway system. Make sure that every tool call request either you are doing this from your local or through agent, it has to go through some sort of centralized MCP gateway. Uh make sure it authenticated before it reaching to the MCP servers. And then make sure that your MCP servers take a control on what tool um you know, uh you should use uh to execute a particular query, right?

And make sure that you have monitoring in place and complete auditing in place for that tool executions. And of course, you should um you know, keep the always human in the loop process for every um every action items, every high-risk items. Um make sure that there is a observability um and you know, um auditing has to be in place. And last but not least, make sure that context optimizer should always be there to justify your cost on your agents, right? So, just want to share is that I released my MCP book recently on Amazon. So, feel free to download. It's free in Kindle. And it's hard copy is also available. All right. Thank you so much for watching this video. And if you do have any questions, feel free to, you know, like reach out to me over LinkedIn or Twitter. I'm happy to answer. Thank you again.