# How to avoid disaster when vibe-coding a billing engine — Andrew Garvin, Stripe AI Engineer · 2026-08-28 Andrew Garvin, co-founder of Metronome, the usage-billing platform Stripe acquired this year in its largest deal ever, argues that billing engines carry deep business logic, so a coding agent should build a test environment while a human stays in the loop — demonstrated by a Stripe Projects workflow replicating Lovable's pricing model. He demonstrates a billing sandbox that provisions a customer, metered usage, scoped credit pools for builds, plan mode, cloud and gateway calls, and a draft invoice, using portable skills files and verbose errors so the agent can self-correct. He separates agent as product, buyer, and user, pointing to HubSpot's move from seats to credits as one agent does the work of many logins. Stripe's CLI use by coding agents has risen exponentially. ## Questions this episode answers ### Why does Andrew Garvin recommend keeping a human in the loop when using an agent to build a billing engine? Garvin says a billing engine is business critical and carries deep business logic, so the goal is not for a customer to operate the entire system without a human in the loop. Instead, the coding agent should accelerate work and get you into a test mode or test environment, not push the setup into production unattended. [7:13](https://aiengineer.podhood.com/2b885ec0-95de-4012-9f8e-133615e35531?t=433000) ### How do you set up a billing engine that mimics Lovable's pricing using Stripe Projects and Metronome? Andrew Garvin enters a natural-language prompt asking the Metronome billing agent to create a demo billing engine in Metronome replicating Lovable's pricing model. Stripe Projects provisions a Stripe account and the Metronome instance through the CLI. Portable skills files give the agent context, and it flows usage so you can test the setup in a sandbox before production. [5:31](https://aiengineer.podhood.com/2b885ec0-95de-4012-9f8e-133615e35531?t=331000) ### What is the difference between an agent as a product, an agent as a buyer, and an agent as a user? Garvin breaks down three roles. An agent as a product needs usage-based pricing because it can run up a token bill. An agent as a buyer uses Stripe Projects CLI to procure a Stripe instance and backend services. An agent as a user is why HubSpot is moving from seats to credits, since seat-level access to a product is no longer important. [9:45](https://aiengineer.podhood.com/2b885ec0-95de-4012-9f8e-133615e35531?t=585000) ### What did the draft invoice in the Metronome demo show for Lovable's pricing? The draft invoice mirrored Lovable's credits-only pricing model with monthly auto recharge and multiple scoped credit pools. Garvin clicks into build credits, plan mode credits, cloud credits, and AI gateway credits, with usage drawn down from a first-class credit object; overspending produces an invoice at the end of the period. [14:22](https://aiengineer.podhood.com/2b885ec0-95de-4012-9f8e-133615e35531?t=862000) ## Key moments - **[0:00] Intro** - [0:13] Stripe acquired Metronome in its largest deal ever, says Metronome co-founder Andrew Garvin - **[2:29] Billing risks** - [2:30] Metronome metered OpenAI and Anthropic before they had any revenue — and now gives agents wallets so spend can't run away, says Andrew Garvin - **[3:40] Live setup** - **[6:02] Skills files** - [6:07] Metronome's portable skills files give coding agents the API's hard-won context so they don't hit billing footguns, says Andrew Garvin - [6:47] Metronome writes deliberately verbose error messages so a coding agent can self-correct during setup, says Andrew Garvin - **[7:13] Human in loop** - [7:27] Don't let a coding agent operate a billing engine in production — Andrew Garvin recommends using agents to build test sandboxes instead - [7:54] Metronome's skills files make the agent stream test usage into the sandbox so you see a live customer, says Andrew Garvin - [9:07] Andrew Garvin says Stripe has seen exponential growth in new business formation and in customers using Stripe and Metronome through coding agents - **[9:30] Agent roles** - [9:45] Andrew Garvin's framework separates "agent as your product" and "agent as your buyer" — usage-based pricing applies to both - [10:37] Andrew Garvin: the "agent as user" case is HubSpot moving from seats to credits — it cut EMEA seat prices because one agent replaces many logins - **[10:38] Seats to credits** - [11:21] All five companies at Andreessen Horowitz's demo day were sales-led agents for SAP or invoicing, says Andrew Garvin — value accrues to one agent user - [12:22] Enterprise coding agents like Cognition, Cursor, OpenAI and Anthropic are adopting CSP-style prepaid and postpaid commitments, says Andrew Garvin - **[13:01] Agent demo** - **[14:12] Credit pools** - [14:22] The agent-built Metronome sandbox shows a draft invoice with Lovable-style scoped credits for builds, plan mode, cloud and AI gateway, says Andrew Garvin - **[16:27] Beyond billing** ## Speakers - **Andrew Garvin** (guest) ## Topics Vibe Coding, Coding Agents, Developer Productivity ## Mentioned Anthropic (company), Cognition (company), HubSpot (company), Hugging Face (company), Lovable (company), Metronome (company), OpenAI (company), Salesforce (company), Stripe (company), Vercel (company), Claude (product), Cursor (product), Stripe projects (product) ## Transcript ### Intro **Andrew Garvin** [0:13] Hi everyone. Uh, my name is Andrew Garvin, and I'm one of the co-founders of Metronome. Metronome, if you're not familiar, is the top platform for usage billing, which, as you might imagine, is taking offright now. Um, so much so that earlier this year we were acquired by Stripe in the largest deal that Stripe has ever done. Um, and what I'm going to show you today is a fun project that we've been cooking on with the Stripe team, um, that is a demonstration of some of the things that we could work on together. Um, this is going to combine two very different, uh, products within Stripe. Um, if you haven't seen it already, I recommend looking at Stripe projects. Just o-open up, open that up on your phone while I'm going through this. Um, this is something that actually was launched literally the week that Metronome was acquired, and so, as you might imagine, the velocity of development at Stripe and at Metronome is pretty high. Basically, what Stripe projects is, is it's an orchestrator to allow you to operate and build your business as fast as possible. Um, and so, in essence, what it does is it provisions a Stripe account for you, as well as backend services that you may need. Think like Vercel, P-Postgres, and in this case, a Metronome billing agent, in order to launch a product quickly or an application quickly, all through the CLI. Um, so the demonstration that we're going to do today is a very simple demonstration of how to get set up through Stripe projects, but you can imagine all the sorts of things that people are building on Stripe today. Um, one of the really cool aspects of being inside Stripe is their scale and data. Um, and one of the things that we've observed inside of Stripe is that the use of Stripe's CLI has exponentially increased over the course of the past five, six months. We're thinking about all of the implications of the coding agents operating systems, which we'll get into, uh, a greater topic on later in this demonstration. Um, so the beginning part was, uh, sort of the thesis for this is how to avoid dis-disaster when vibe-coding a billing engine. Uh, as you might imagine, being in the billing space for multiple years now, we've seen all sorts of crazy things happen, and it's even getting crazier now that people are expecting to operate Metronome, a very complicated and deep product, with a coding agent. Um, and so, as a result, we are working from a developer experience standpoint to make this a more seamless experience for people and to a-ave them avoid disaster. Um, and so we're going to see in a second what that demo w a demonstration of what that might look like and how we're guiding things. ### Billing risks **Andrew Garvin** [2:30] Um, but just to frame it, the what's happening today in with launching agent products can go crazy, and it can go sideways in all sorts of ways. So as just a couple of the types of problems that Metronome helped solve for folks, uh, first, w-we are, for example, and have for many years now, taken in all of the API calls to OpenAI and Anthropic and metered that for those companies. We've worked with those companies since before they had any revenue. Uh, and so, obviously, we've operated at global scale, operating a metering service with a number of different data impacts. In addition to that, we also operate credit models. Obviously, today, with usage-based pricing, it's not just about having a pay-go, uh, metered business model, but also all sorts of different forms of credits, commits, sales-led, uh, discounts and offers. And then finally, now, in the in the last six months especially, the impact of failures here is growing in importance, in particular because agents can run away with spend. And so we're thinking about how to give more, uh, controls to our customers that they can offer to theirs. So think like, for example, having agents have a wallet, um, that they only they can spend from, and having controls at that level. Okay. So that's what we're going to go into today. Let's actually get into doing this demonstration. You're going to see how simple it is. Um, so I'm going to initialize Stripe projectsright now. ### Live setup **Andrew Garvin** [3:51] Uh, oops. And this is, in fact, a live demo, so you should expect all sorts of different things to happen here. So what this is doing is, again, initializing Stripe projects. Um, we are going to I think this should work. Um, so we've selected that we're going to use Claude here. And we have a very simple thing. Wh-what I, I go around the world basically talking with companies about their business model. It's not initialized. **Guest** [4:17] Space. **Andrew Garvin** [4:18] Uh-huh. I see. Allright. C-come on up. **Guest** [4:27] That's a claw. **Andrew Garvin** [4:37] Go. I got it. **Guest** [4:38] Let's try it. **Andrew Garvin** [4:38] Allright. Thanks, Dan. **Guest** [4:39] Space dash five. **Andrew Garvin** [4:47] Allright. Sweet. Um, yes. Let's proceed. Cool. Allright. So our product or our project is, in fact, ready now. And so I go around the world talking with companies about how to set up their businesses. One of the things that's really top of mindright now is, um, replicating certain business models. People want to get off the ground without having to think about it too deeply. One of the key topicsright now is replicating Lovable's pricing model. And so I'm going to enter a prompt in natural language that allows us to guide the Metronome billing agent and Stripe to set up a demo account that has that, um, that has that element to it. And create a demo billing engine, uh, in Metronome, mimicking the Lovable pricing model. Oops. Just make sure I don't misspell that. Um, and so if you want to, on the side, you can see you can open up Lovable's pricings, and you can see all the different elements to it. But, uh, in particular, they have a prepaid credit auto recharge model, which is very common in, um, in sort of a self-serve motion today. Um, so let's get this going. Um, so while this is going, I want you to pay attention to a couple of things that will happen. ### Skills files **Andrew Garvin** [6:07] Um, and in particular, if you look at the obviously, the slide on the left, there's a couple of different points that I want to hit on. So first, um, again, Metronome is a very complicated and deep product, and, uh, there's a lot of different ways to hit footguns, etc., uh, if you're not guided. And so what we've invested in is building an extensible set of skills files that can provide context to the agent that's implementing Metronome and working with our API. Um, these skills files are also portable and easy to install, so you can use them on your own side. Um, and what it does is it allows us to essentially remove the friction associated with getting started. Um, and so here, uh, that and that's very important because you want to be able to test and work with the product and evolve over time. You're seeing here an error code, um, from a developer experience standpoint. You know, this is nothing new, but our per our perspective is to have much more verbose and clear errors so that the agent can self-correct. Um, and, uh, and so, again, our developer experience teams are working on finding more failure cases like that, um, and being able to help guide, especially in the initialization and setup. Um, one thing that's, that's important here, uh, I think keying off of the last talk that was in this room, um, the goal that we have from a product development standpoint is not to have a customer operate the entire system without a human in the loop. ### Human in loop **Andrew Garvin** [7:27] This is a type of system that is both business critical, has deep business logic behind it. And so instead, what we are recommending and building toward is to use the use your coding agent as a way to accelerate your work and get into a test mode and test environment. Um, and so, again, what we're doing here, we're not expecting to ship into production. We're not pushing it into production. In fact, wh-when we go into the Metronome environment, you'll see that basically what we've done is built a sandbox experience. Um, and in the Metronome context, there's what it means to sort of test your initial setup is not just that you can see, uh, a contract or something like that, or see a customer provision, but also you need to see usage. And so in be in on the backend here, our skills files are directing, um, the agent to actually flow usage into the Metronome platform so that you can see what a live, uh, in li what a live customer would look like. Um, I'm going to do one more beat on Stripe projects here. I think in this case, what's happening is that Stripe projects so, so using Stripe CLI, it's, it's engaging with Metronome, which is an external vendor here. You could also imagine also provisioning a bunch of other applications, uh, and using natural language to call for those as well. Um, and so what again, what's nice about this is that it basically removes the friction associated with setting up a test application or a test environment. Um, and we, again, have seen a ton of, uh, of usage of this form, um, and we expect to see even more. Um, this is sort of like from our perspective coming into Stripe, this is one of the things that's been really amazing is that Stripe is on the forefront of thinking about agentic commerce and preparing primitives for the moment that we're inright now where, in fact, this is exactly what's happening. Companies that are launching new, um, new applications and new businesses, they we've seen an exponential increase in new business formation at Stripe. Um, and then in addition to that, an exponential increase in customers that are using Stripe and using Metronome through the coding agents themselves. Okay. Um, we are almost done here, I believe. Um, while this is going, I, I like another sort of like thing to sit back and think about. When I go around to product teams today, they're obviously thinking about building for agents, but I think one of the things that's important to do is to de-decode what exactly does that mean. ### Agent roles **Andrew Garvin** [9:45] And so I what I like about this sort of framework for thinking about the coding agents today is thinking about the different roles that they play. So obviously, companies are launching agents as a product, and therefore, that's one of the reasons why they need to have a usage-based pricing model because if the agent can be the product and run up token uh, token bill, it's important for you to be able to meter on that. Um, what we're talking about here with the Stri-Stripe project CLI is the agent as a buyer. So literally procuring their initial Stripe instance as well as, uh, additional backend services. That's important to basically make your services discoverable to agents that may be building an application or working in the open web. Um, on the Stripe side, that means both in a B2C environment, so they're working on agentic commerce, but then in the Metronome environment, we're talking about in a B2B context. Um, and, uh, and so there's sort of like multiple different levels to play out there. And then finally, one of the reasons why Metronome is really taking offright now is because of the agents as a user. Um, and so, for example, we've been working with HubSpot for the past couple of years. They are currently on a path to transform their entire business from a seats-based model to a credits-based model. ### Seats to credits **Andrew Garvin** [10:51] Um, if you've seen some things in the news, that's starting in, in EMEA, where they have dramatically lowered their seats-based price and added on a credits-based model. The fundamental reason behind that is because what they need to be concerned about is a world in which an agent can operate their entire system. Um, and in that world, essentially paying for a seat-level access to the product to perform your work is no longer important in some sense. Um, we've been talking about this in, uh, as sort of headlessness. You know, Salesforce, various others have ta have talked about this. This is what Metronome is literally seeing today. Um, and so as like one example, last week, I was at a, uh I was at, uh, Andreessen's demo day where all five of the demoing companies were sales-led agents meant to operate platforms like SAP or o-operate, um, operate invoicing platforms, etc., etc. Um, and again, in that world, it's important for you to have a usage-based pricing model because you have the possibility of essentially all of the value accruing to essentially one user of your platform, which in this case would be an agent. If you guys are thinking about pricing models and not just developing in the agentic space, I'm a good person to talk to. I'll be out here in a second. Some of the things that we're, um, thinking about here are basically, um, not only having a credits-based model, which has been, uh, dominant on the market since OpenAI launched their prepaid credit auto recharge model a couple of years ago through Metronome, but in addition to that, o-offering more and like more and extended offers, including in a sales-led and an enterprise environment. Um, and so, for example, what's happening with all of the coding agents in the enterprise, think like Cognition or Cursor or OpenAI and Anthropic themselves, is that they are starting to adopt more, um, uh, commit structures like the CSPs have done for the past 10 years. Think having prepaid commitments, postpaid commitments, and specific types of offers for specific types of customers. Okay. I think that we should be good to go now. And let's see what it looks like when we open up Metronome. ### Agent demo **Andrew Garvin** [13:02] Okay. What's up? Uh-huh. We've done multiple different versions of this, as you might imagine. Okay. So as we open this up, uh, pull this. It's fun. I love all these demos where you're just looking at people logging in. Um, so as we open this up, so the general pain that Metronome has is an onboarding wizard meant for a human that needs to, uh, set up their environment. We've we don't need this now because we had an agent set up this environment. And as I come in, you're going to see some of the core Metronome primitives here. Let's start by looking at the customer that was set up. Again, this is for testing purposes. Um, up at the top, you're seeing the customer with a certain lifetime spend. This was auto, again, populated by, um, by the agent for the demo environment. Um, I'm immediately going to go into their invoice, and we'll come back to this in a second. ### Credit pools **Andrew Garvin** [14:22] Um, and so what you s what you see here is a draft invoice that was created associated with, um, sort of replicating the Lovable pricing model again. Um, so if you have this on the side, you can see all the different elements of that. But the core aspect of Lovable's pricing model is a credits-only pricing model where you auto recharge on a monthly basis. Um, and then in addition in addition to that, they have multiple different types of credits that are scoped to diff different types of usage. Beyond the, the use of those credits then, if you go over and sp if you overspend, then you have an invoice at the end of the period. Um, so a couple of the different like concepts there that are relatively complicated to administer is the credit itself. And so Metronome has a first-class, uh, credit object. Here, what you're seeing is that there was a crea credit created for the initial period that we're testing for. We had usage that, that draw that drew down from that entire credit balance. And then finally, if we go back to the customer pain, um, in addition to that, you can see the usage that we that we plopped in. All obviously, in a production environment, you would be seeing this in against real usage that you have. The core reason, again, to show it in this manner is to just see what it would look like if you adopted the pricing model and then had real usage against it. Again, I'm going to come back to the invoice. And so here, you can click into each of these different components, um, build credits, plan mode credits, cloud credits, AI gateway credits. This is exactly what the Lovable pricing model looks like. And again, the way that we coached the agent to be able to do to, to build this was just describing a natural language to replicate Lovable's pricing model. There's nothing more difficult than that. Um, so without going into Metronome's platform to, to great an extent, um, the what we just did here was we in-initialized, uh, and created a Stripe instance. We then init through Stripe projects, uh, we also created a Metronome instance. Then we coached the agent to be able to, uh, build a demo instance of Metronome that had a real pricing model live in production. And so you could imagine basically testing then from there the exact testing and tweaking from there exactly what you wanted before bringing that into production. Um, this sort of framework for thinking about development both applies to Stripe where we are working very, very hard to make it easier to, uh, run a complicated business model and get off the ground, but also I think it bears lessons for how we might pursue agentic development more generally outside of Stripe. ### Beyond billing **Andrew Garvin** [16:42] So again, think about some of the primitives primitives that we talked about here today. Agent as a buyer, agent as your product, agent as your user, and disambiguating what the different the different, um, modes and, uh, and like implications of those are. And then in addition to that, um, having d having ways in which we coach the agent to operate more effectively in including in a in a in a difficult environment. You can try this for yourself now. So, uh, the easiest way to get started is with the commands that are that are listed here. Uh, and you can see everything that is available through Stripe projects through Stripe projects online. Um, as there are a number of different providers that are onboarding every day. Um, so companies like Vercel, um, like Hugging Face, etc., are basically like working in Stripe projects environment to be able to make their own products more discoverable to agents that are operating Stripe's system. --- This library is powered by PodHood (https://podhood.com), the podcast website platform.