Open Source Friday with Kong's OpenMeter

Hey, hey, hey, hey, hey. everybody. Welcome to Open Source Friday, Tasja. Thank you for being here. Thank you. Thank you for having me. It's an honor to be here. Perfect. So, even though we are coming to you on a pre-recorded stream due to travel and life, we are both Deja and myself monitoring the chat right now. So throughout this conversation, if you have questions about the project, if you have questions for Tij or myself, feel free to drop them in the chat and we'll be sure to get to them. But before we get started, let's get to know you a little better. Could you just briefly introduce yourself? What's your current role and how did you get into Devril because I don't get to have a lot of Devril people in this show. So this is interesting. So yeah, I work for Kong. I'm a product advocate comes under the kind of developer relations and everything but I belong to the growth team at Kong and I look after how do we enable our developers to start actuating the product and bring closer to what they can start using the product better. So any sort of things that they wanted to get implemented get they get stuck is something we been the first point of contact and bring it to them implemented. So all that we care about the developer experience care about their enablement and they should be able to have something running of what they're looking for. So whatever it takes whether it takes writing whether it's right recording whether it's right writing code whether it's right recording as we're talking about is stream whatever let's do it. Okay amazing so for people who are new to Kong including myself what problem does Kong solve? Where does it sit in this sort of the new platform stack? Can you tell us a little bit about the background of the company and then we can get into the project? So Kong is an API gateway as how it's known to the market but essentially Kong is the developer of this cloud API platform or the AI connectivity that it's currently. So whenever we wanted to have the services connected throughout it's essentially the core gateway that stands in between able to completely manage your APIs govern it secure it and then you process all this through one single centralized point of gateway. So that's essentially what Kong gateway is all about. It's something started as an open source project and then it grew out all in these many ways essentially now even now at the part of like when we have AI come in and the AI agent take development and people are fully in AI development. So Kong places the position in terms of how you can actually stream your intelligence through APIs because even if you go and see about Kong they talk about very clearly there is no AI without API and you got to put APIs somewhere and then stream it. So Kong is that product and that gateway that stands in between helps you serve whether it's data whether it's intelligence whether it's any sort of services that you're running in your own systems that's a brief about what Kong is and it has its own sort of offerings whether it's AI gateway API gateway mesh and even gateway developer portal ming and billing is something that we'll be focusing on today which is something developed underneath another open source project which is called the open meter so the open meter is the project that what we're diving in is that aspect of Kong metering and building. So yeah, that's pretty much about what Kong is and what we would be diving in. I'm interested in open meter then can you tell us what problem was trying to be solved by creating open meter and then would love to see a demo. Absolutely. So right now where Kong is positioned it's been the AI gateway as a service. It has the initial the conjecture of things coming in together and then you kind of manage and orchestrate or even do just secure and process all together with that plug-in kind of an architecture for the API and AI gateway. There is something that when we look about you have your own API now and you wanted to go forward and actually put it for your developers. So you have your old APIs and you wanted to make it public or for your partners or for anybody you wanted to have that developer portal. So that's where Kong have the feature. It's all it's all about developer portal. How you can expose your APIs to a portal and then let developers you serve your developers through the APIs. That's what the developer portal about. But now when you want to start charging or even monetize your APIs or even the AI services that you're running in your Kong or even externally wherever you're having your own services, if you want to start charging them mainly very specifically about usage based billing. So now you want to charge your APIs very particularly based upon their usage. So that's where the Kong metering and billing comes in where it can integrate with your applications or internal Kong services and then start metering. So when I say metering, think of it like an an electricity, right? You start using your electricity and then you start metering it. So then at the end of a certain month, you'll get the bill. So you'll completely calculate and precisely see the consumption and what called as usage based and now you'll be charged accordingly based upon your usage or consumption. So it's pretty much the same but in the world of APIs and the software now you have a metering service right there depending upon your plans subscriptions features rate card and everything al together it will actually monitor and completely track and iterate and then bill it. So the problem is the developers when you go at scale doing it by yourself is something turning out so difficult. Imagine developers are doing it and as you go scale up things change. For a simple example even for the AI models maybe chart GPT or even for you know anthropic who are the models they offer the amount of tokens that we pay for input versus output is the output tokens really are more expensive than the input tokens right because the processing being there and the amount that it keeps you know sends you out would be little more expensive definitely than the input. You just have a flat rate or you have rate limits or you just have seed based. Now it's all completely usage based and particularly how you track them, meter it and then start building. So that's the problem we want to solve together. That's where open meter came in. It's 2023 project. Peter and another co-founder together started the project. Anders Anders and Peter started the project and then it was part of the Y cominator as well and then slowly moving on forward it has built up and the founder is also from the background of stripe and basically the problem has come through focusing on infrastructure casting infrastructure metering I think I could share it so they can learn more about the source of how it began and everything the problem that was starting to bring open meter come live so that's the backstory about what open meter and all about but then recently September 2025 open meter became part of Kong and now Konging nurturing the whole open meter project and now open meter is also a product part of the Kong connect kong connect is a cloud version of the open source API gateway and the open source open meter now it's part of the connect where you can start using and just maybe we'll look at a demo how we can start expo exploiting open meter lively in the cloud so yeah that's pretty much where it all began and how's it going and the problem Very cool. Okay. Oh, so it became it was an acquisition by Kong. All right. Well, let's take a look. Let's see this in action. Absolutely. Okay. So, what you see on the screen is connect. This is called connect here. So, this is pretty much the cloud version of the Kong. So, I just signed up. Maybe you can also sign up if you are. So, I'm setting up the probably use case of how we want to go by it so that we can explore what it does. So, the first point imagine we have an our own API and I'm setting it to a particular gateway. So I'm using a gateway, you know, passing and making it as a proxy. I'm going to use my API gateway as a proxy to access my APIs. I'm going to have my own customers and for those customers, I'm going to meter them based upon their usage. So for every call, I'm going to charge a dollar maybe and then start and send them an invoice. So this is usually the use case, a very bare minimum use case simply to show how APS can be started to monetize. So that's the use case. So let me see if I have a chart to you know bring out. So this could be what we could do simply. So you see that we're going to put out a gateway setup and then maybe just show about how you can set up the metering catalog and then you're going to configure your own plans right when you start to monetize you got to put out what features you're charging for and what plans and what subscriptions and then at the end we'll just validate like I'm going to make at this phase we're going to go ahead make a few calls and actually see it's starting to meter and it's starting to give us an invoice and starting to build for all the people who are exploring your own kind of metering this could be a good start so you can just look out So initially let me go ahead and actually quick start you know setting up an API gateway so that it would be really quick. You can just click on new gateway from there and you can choose serverless for the demo purpose or you can go for self-managed if you want to go by you know your own thing of docker or linux binary or kubernetes but for the demo purpose we can stick to serverless to keep it simple and then I'm going to call it as like an example service I'm just configuring a gateway right here and it just creates gateway on the go and there it is so the gateway is created the name is that example service so next I'm going to quickly set up a simple route And the API I'm going to use is a flights sample demo API. So I just say get flight details and this is the API. This is my real API. So you must imagine this as your API that you're going to charge. So if you see the URL like api.com/flights. So this is the actual API and now we're going to put it behind an API gateway. So that's what we we are doing. So I'm just giving up the details and the route and the route path. So this path is going to be the path for your API gateway to access this particular route. So let me save this and yeah it's that quick. You can just quickly because it's cloud and the serverless approach. So now this is the proxy URL which is also the gateway and then this path is what takes you to that particular endpoint which we have just seen. I can quickly copy that just open that up in a new tab right up there. So we can just see the same data coming up here. So it's because of the serverless setup, it was quick. And then the gateway is now proxying your API. And that's how you quickly set up the gateway. And if I just do a refresh, you should be also be able to see the request should be count changing to one or two. So yeah, you're done creating a gateway. Now we're going to go ahead and actually go for the next step worries about you're going to put out some authentication. Let's quickly put out some, you know, key authentication. So that that's how you charge rate. Even if you wanted to charge a particular user, you need to have a customer and their own token. And based upon their token, you have to track how much they are utilizing. For that, we need to do two and three which is about enabling an authentication and putting out the route strategies. So I'm going to go that going to be done by you know you can just click on add key authentication here or we can go to plugins and just add a new plugin and the plug-in we're going to do is the key authentication and doing it for scoped. I'm going to select it for only for the Kong air flight sample API and I'm opening it for all the routes. So yeah, we just created the key authentication plugin. So now the API expects an authentication thing to happen. So that's how you are enabling an authentication key authentication to happen. But we need to know who is accessing. So for that to happen we need to have consumers. So who consumers or customers like similar terms right? So let's go ahead create a new customer and let me call it like a Kong air. So simply just saving it and for this particular consumer I'm going to give some credentials. I'm doing it all manual. you it's best done in a real time with APIs or SDKs and calls together when a signup happens and all that. What I'm showing is a demo where we have a consumer manually created and I'm manually assigning a key authentication for them. So this is the key. So I'm just manually giving a key which is going to be a credential for accessing for this particular customer. So just save it and that's it. So you have created a consumer called cong air and you added an authentication layer which is like a key authentication and the key is so yeah that's pretty much we are clear with setting up customer enabling authentication. Now let's get into the actual metering and part so you have the this is all what we want for quickly setting up what do we call the boiler plate maybe maybe I'll just send a quick request and show so that we are all set. I'm going to copy this proxy URL and then routes. So this is the path the first path. So I'm going to send by the way what we are also using here is called insomnia is also an open source project which is backed by com product is insomnia. So now if you see if I try not to make a request it says no API key found because there is an authentication layer behind. So I'm going to go ahead to go to o keep like API key and then the key is going to be API key and the value is add key. Okay. So if I just do and send a request I'm getting the data. So let's go and start adding metering services on top of it for that particular customer. So for that we'll get into metering and building. And the first thing we need to do is to enable particular gateway. So this is how we do. You just need to go enable a gateway. And the gateway that we're going to do is an example service that we just created. So you can just quickly enable it. And once you've done it, next part is mainly about what features. So now we need to come back to talk about the whole thing of pricing. So if you see superbase pricing just for the sake of explaining what all these are because we're talking majoring and billing now we have a free version we have a pro version. So all these are like plans you have a free plan pro plan and team plan and each and everything is a feature feature by itself and you have entitlements and you have certain the usage base if you see there is and on default you have you know 100,000 and then from that from next you know based on your using your usage you'll be priced. So there are rate cards, there is entitlements and there is this is all features and plans and then once you have it it's a subscription by itself. So with all this kind of little terminology to be familiar with now it would be more convenient for us to relate. So what we going to talk we're going to create a new feature first so that that allows us to do the feature. Let's talk about tokens. So based upon tokens we are going to charge. So I'm going to create a feature called token for example and then the meter is going to be about API gateway request. If you're doing AI gateway kind of a stuff you can choose that meter because it's a API gateway. We are just doing an example with tokens. So we're going to choose API gateway. And then you need to add a group filter because I'm going to filter by the service. So the service we created is an example service. So we have to filter that out uh among all. So based upon the service name which is equals to it should be the example service or the Kong Kong service. So it's API gateway it should be autopop populated then here it should be I don't know why it's not coming but it should ideally because it's a live demo. [laughter] So generally yeah this is example service and then you have gateway services. So this is the one what we need to add there. So let me do it manually then if it doesn't autoop populate. But while you're cranking it up there, I did want to highlight also what you mentioned that open meter has been built as a through open source project. So what you're seeing here being the the cloud version, but you can self-host this if you want. So and then figure it out if if the cloud version is for you. But you know, I haven't thought about how many people are probably undercharging for usage. So this is very timely. [gasps] Totally. Yeah. And mainly at a time like this, many founders are coming out building their own AI products. They need to start building more precisely, more accurately for the usage. So that's where it comes in. So when they have they can set up open locally. But the thing is the biggest miss is the UI because developers they go by config YAML and everything. So the open source supported the UI is a big part of what we're doing right now is because of the cloud thing which has few fields to set up conf automatically. But if you start using setup by open source go you have to deal with the YAML you know config files and then start away. So yeah it should ideally this is the thing but I'm taking a bet so I may come back and change this. So I just created a new feature which is about token and based upon this feature we'll be starting to charge. So that was about feature [snorts] and the next is about we need to create a plan. So now you have a feature now you need to create a plan on top of it right. So let me create a new plan and I'm going to call this plan as a premium plan. So just the sake of it so I calling as a premium and the billing cadency is keeping it as 1 month and just the currency is USD and that should be enough. And in the same plan you should also have the rate cards like how much do you want to charge for units. So that's where we come to the concept of rate cards. So I need to click on just add a rate card and u the feature which I'm going to add a rate card is for the tokens and next is about are you going to make it like a free or a flat fee or usage based tired and all of those things. For now for the demo let's use usage based I'm going to charge based upon how much they consume. So monthly and per unit I'm going to charge $1. So if you see the feature is token and based upon I'm going to charge $1 per unit and that goes for a month and not putting any entitlements and then just save the date card. So I just created a feature and a plan and a plan includes the feature and the rate how much you want to charge for it. So this is pretty much how you can quickly do it and you just need to publish the plan. So you have a button here. So you have to once you publish it it will be live. I mean once you have these things configured about your site, your pricing pages and everything that can be automated. So then now we're going to go do something that unusually done by UI creating a customer. So I'm going to create a new customer and map it to the person which we had at the top. So I just selected the Kong air which we just did in the AI gateway. So Kong air and the Kong air I'll just use it as it is and yeah that's pretty much you can just save it. Then so the customer I just created a new customer and for that customer I'm going to add a subscription. So imagine I'm the customer I got into here to your account for your APIs and I'm subscribing for something. So what am I subscribing for? I'm subscribing for the premium plan and it would be starting immediately and it's going to bill on a monthly basis. So this will be going up like $1 per unit and then that should be pretty much. So this is about $1 per unit and you can change either. So I can just start the subscription. So as I click it, my subscription is now started. So just that now if I use my token and start calling this, it should it will be starting to charge. So if I go to my billing and so this particular customer and their subscription is about premium and they're invoicing. So right now the invoice is $0. Now if I go start making some request so if I just do a quick request 1 2 3 4 five. There you go. You see the five there because I made five request. It's five. So if I just do 6 7 8 9 and 10. So you look at the thing now it's relating that with the token identifying the customer and taking it and having a complete obser you know observation or this tracking of the usage and adding the dollar value that you put to the rate card and immediately showing you the invoice or how much it charged. So it's now $10 because I made 10 request. So then you can quickly you know preview invoice and you'll get the invoice about it. And then from here you can configure your stripe payments or you know HubSpot or whatever you have and then collect the billing. That's pretty much how quickly you can simply connect and start billing your users and monetizing your APIs or your AI agents. I like that. So how are people doing this now? Like at the application level they just kind of I don't know what do you do like a database and a prayer and hope [laughter] and hope that that it works. No, but honestly, how are most people for the developers that you're talking to? Is there a lot of people that are still trying to track usage inside their application code? And then it sounds like a ticking bomb to have it set up like that, especially now where if it's an API for anything related to aging work where one call can be 50 all of a sudden. It sounds like it would be very challenging. So this is obviously meter tracking for usage is not a new concept but the way to track it as part of the infrastructure to track it. I think this is very novel and that's awesome actually adopting this level of awareness of what they're charging. Can you tell me a little bit about some of the experiences that they had? Definitely. I would think that that will free up a lot of the space if you're tracking it in your application itself or if you're using another paid thing to try and track usage. If you have an example of someone that or like either would it be a project that's open or one of the customers that you have that is like they were using this they weren't using this for tracking and now they're using this and like just like a quick example of like how I guess this is kind of like improved that bottleneck so that people are not like disputing charges or like worry about their bills because they know they're charging what they're supposed to be charging kind of thing. One quick example is simply about now that one simple thing to you know most straightforward not to keep anything complex. So where you have a use case of having an internal usage of multiple LM models. So you now have an you you use cloud, you use chat GP and you have publicity and Gemini and all of together when you start giving access to your own users you would essentially wanted to have some sort of a complete control over the billing at the same time a complete visibility over the billing at the same time you also wanted to have a big picture of what being consumed of how much and maybe it's like a dashboard of looking out which models are being consumed the most and what sort of gods that you want to put for each and every models And although now you have a product now you wanted to integrate with a single endpoint that allows you to access chart GPT or cloud by just changing the route or just by changing some header values. I mean that is something that's done by the API gateway or the AI gateway which is the base and the flagship open source project that Kong offers for the API world. Adding on top of that now you have this thing configured. You would be able to actually track it and essentially just give out and bill people accordingly based upon their usage. That's quickly one of the simplest use case that you can quickly also see for most of the products and now people are just having when they have their own AI product it involves various models and above all they have their own models too they have their own charges their own pricing and everything together. So what we have seen is more in terms of connecting with an API gateway from Kong to the metering and billing product in Kong. But if we have our own uh product and we have our own endpoints and services running open meter has an SDK so you can actually infuse the SDK into your code and then start making calls then the data can be ingested into the cloud version or your host version. So you can just track it all together. So that's how you can also go by as you starting to implement this ming and billing for your services and start monetizing your AI agents or monetizing your API endpoints and all of it together. So listen I'm here for a team get paid. So this is a quick way for you to put that measurement layer into everything that you're exposing and it tracks a lot of stuff right like not just you just gave a really good example for the model API for the model usage but like anything data pipelines anything like if there is anything being consumed you can track it and you can make sure that you're charging for it so this is brilliant okay so if I'm watching this and I'm like super into this this metering life is my life how people can get involved with the project like I love for us to share a little bit about the open source side of it and then anything else that you want to show us. Yeah, totally. So we have the repository right in the you can just open up any sort of an issue that you have. So you have let me open that up. So you have open meter and open meter is URL. So once you go up there you have everything you can raise a pull request for any changes or you have issues running up there you feel free to pick them up but essentially we have we pay some attention now it's been taken care completely by the same open meter team is part of Kong now so they have the closed eyes on it so for any contributions that you can just look so I can just add that link so you can if you go to kong.com/ community so you would be able to find a link here for joining the slack so you can just use this button and once you click on that you'll be able to join the Slack community where you can talk about maybe if you want to try the product try implementing or you just wanted to just go ahead and just start using or you know what do you call so you can join Slack hit me up say hi Tja and you just say you want to do something so you can just come to me and I'm happy to help to build you anything or you want to contribute I'm way too excited come and contribute to the project you can just open up if we find any issues pick it up or you create issues we have docu documentation as well. If you want to contribute there, you're most welcome. We recently have a blog published stating how I built my AI agent and start monetizing it. So yes, let's let's share that link to the blog post, please. That's something that you can actually start and see and start using it. So we have documentation as well. So if you want to contribute to our docs as well, you're most welcome. So there is always some work to be done or know to be have an opportunity to actually show out and actually start contributing to docs or content or anyways. So you're most welcome you come and start doing it. If you have ideas do let us know in the slack as well. So that's how you can get started and that's also some of the easiest ways that you can just quickly start doing things. So I'm more excited to help you or guide you anyway needed. I'm also sharing the link in the chat for the blog if you want to try out. And so this was your own experimentation. Can you tell us a little bit about what you did there in closing because I think like that's one of the funnest part of this. So can you tell us a little bit about what you were trying to fix how you did it? Yeah, absolutely. So I have this use case that I have my own AI agent. So that's has its own kind of service where it's want to do translation or it's a voice AI and doing any sort of services that you're now hearing a lot about the AI products coming in and I'm sure most of us when we jump in start wanting to build something with help of AI in behind there's any number of use cases that you can think of and you know there's a lot already booming and people started earning a lot for me I wanted to integrate that metering billing and how you start collecting payments the biggest point which I already revealed is what in the blog already which is about the input tokens and output tokens and how you calculate is a difficult part. So that thing is something what wanted me to get started and actually write about it when when doing with metering and billing with the AI gateway. There are solutions talking about how you start metering and billing but imagine you should be able to track the tokens track the request right at the gateway level. So Kong being an open source gateway provider letting it actually track your tokens, track your API requests and everything and now you have open meter attached to it where you can actually start metering and billing it. So you have the best of API gateway and AI gateway and you have the best of metering and billing all together will what gets you something a more beautiful and more effective and precise billing for your users. All you need to do is just to integrate with some sort of payment provider and start using it. I genuinely have not found this solution anywhere because you have gateway itself to handle the traffic when it go scales up and the same gateway is now allowing you to completely track it and just now there is a billing service that allows you to send bills too. So that's all the blog is about. You can just quickly get started to do all of this. Amazing. Okay. Well, I look forward for all of you to go ahead and try it out and start charging and using an open source project to do it. So that's fantastic. Let me put one more time. I want to share the link to the cloud version. Of course, for folks who are watching and they're like, "Listen, I want the UI. I want it done for me. Need to put a credit card to sign up, which I love companies do that because you can go in and start testing it out, figure out if this is what you want to do." Amazing. Ta. Thank you so much. For folks who want to follow along, they can go on Slack, join the Slack, it will be you helping them, and then take a look at the repository. Make sure if there's anything that you want to actually work on, have a conversation on Slack with TA first before you start working. And then if folks want to reach out to you, is there any other place where you hang out in social or anywhere else where they can? Yeah, I do have LinkedIn. Okay, sure. You can search for it in LinkedIn. I'll be there. So that's all. You can hit me up and and I'd be happy to respond. Yep. Deja, thank you so much for bringing open meter and teaching us about Kong and super interesting. a lot when companies acquire open source projects and they keep funding and they keep caring for the community and this is a good example of that. For all of you who have been undercharging for your offerings, go on set it up. Make sure that you go to the repository and start it, please. And if you have questions throughout this conversation, we'll be keeping an eye on the chat. We might have already answered some of them, but if not, feel free to post them. Same thing, if you're watching this post stream, please drop your questions. I'm sure Tja will be happy to answer. Thank you, Taya, for coming by and showing us Open Meter. Thank you. Thank you so much for having me. It's wonderful. It's a great honor and I hope and I'm very glad the demo went smooth. I was anticipating something could come out, but it went smooth this time and I can see the $10 bill already. So, it's for you. I love it. Listen, this is live TV. Okay. Absolutely. Now, if you're interested listening to this and want to contribute it anyways, please do come. And mainly, if you are interested in writing now at a time, you have agents helping you to write more. If you would like to publish through Kong and write about and contribute whether your content or your ideas or use cases or your some sort of some implementation by itself we're happy. So you can just come or message me in LinkedIn or Slack and would be happy to publish what you have done with the projects. Okay. So use the project and an invitation to publish your experimentations by the official project. So cool. Thank you so much Tja. Take care.