Unkey: Globally Distributed API key generation, with cofounder James Perkins

James Perkins from Unkey sits down for a chat with Mike Bifulco about Unkey's unique approach to delivering an API product that provides value to its users out of the gate, and the OSS story that got them here.

Unkey: Globally Distributed API key generation, with cofounder James Perkins

Show Notes


[00:00:00] Mike Bifulco: . Hey friends, Mike from NPI. As you won't hate here chiming in from the future. I messed up during this interview there's about five minutes of this interview at the very beginning where I am using the microphone on my earbuds to record.

[00:00:11] And you can really tell. So I apologize for that in advance. But I promise I did my best to make it better using all the tools available at my. Disposal using AI and audacity and things like that to make the audio a little nicer. It's as good as it can be roughly given the skills that I have, but I just figured I'd let you know.

[00:00:27] The entire interview is not as miserable sounding on my end. As it seems at the beginning. So zip ahead if you need to, but otherwise enjoy the interview and thanks for watching the indoor listening.

[00:00:36] ​

[00:00:37] riverside_mike_bifulco_raw-synced-video-cfr_podcast_recording w_0027: Welcome back to another episode of APIs you won't Hate. My name is Mike Bico. I'm one of the co-founders of APIs You Won't hate. And I'm excited to sit down today for an interview. Stand up today for an interview, whatever, hop on the internet for an interview with my my old pal James Perkins to talk about what he's building.

[00:00:51] And James, it's really nice to have you on the show today. I'm, I'm interested to hear about UN Key and your story there. Thanks a ton for joining us. How are you doing?

[00:00:57] riverside_james_raw-synced-video-cfr_podcast_recording w_0028: Yeah, I'm good. Thanks for having me. It's been [00:01:00] a while since we've talked. Yeah, I'm, I'm happy to be here. Happy to chat about UN key and what we're building and how we're trying to make APIs suck a lot less than they, they do most of the time.

[00:01:10] riverside_mike_bifulco_raw-synced-video-cfr_podcast_recording w_0027: Oh, you found yourself in the right place then I think for that discussion. James, you and I have sort of a serendipitous internet relationship. I feel like every time I turn a corner online you're poking your head around saying, hi. I, I don't remember exactly where we met, but I know we crossed paths a bit while I was at Stripe.

[00:01:23] I. Doing developer relations there and later while you were doing developer relations and now you're starting and building a product. So I, I'd love maybe if you start just giving us, like, the lowdown on your career. How did you find your way to what you're doing now? And then let's talk a little bit about.

[00:01:36] riverside_james_raw-synced-video-cfr_podcast_recording w_0028: Yeah, for sure. So I've been around a really long time. I think if my co-founder listens to this, I'm really sorry, but I have to do this. So I've been around for 16 years. I started way back doing Java development for a few different companies. First one was like a really small stage startup that did online registration.

[00:01:54] So in America, if you buy a car and you get from the dealership. And they give you like a temp tag in the dealership or the [00:02:00] actual tags depending on where you live. That software was built by me and a, and a few people. And then I moved on to things like gambling and the lottery. Did that for a while, did digital banking, then I moved on to FinTech startups.

[00:02:13] Tina CMS was my real break into where people really know me. That's when I started doing developer relations and doing all the sorts of YouTubes and podcasts and all the other things that I've done throughout my career. And then the real thing that everyone really knows me for was Clark which is where I've met a lot of different people across, across different industries.

[00:02:36] I think that's how we actually met.

[00:02:38] riverside_mike_bifulco_raw-synced-video-cfr_podcast_recording w_0027: I think it may be, yeah.

[00:02:39] riverside_james_raw-synced-video-cfr_podcast_recording w_0028: I think you were working on a project with Clark and you had questions and then I think, you know, bouncing around and then, you know, you moved on to, to what you're doing now. And we chatted a bit about that. But yeah, that's most of my career. And then I left Clark in September which is when the funding for Key was basically almost finished at that point.

[00:02:59] [00:03:00] We, we got a few checks in, but we were still finishing up. And then that was when I went full time.

[00:03:04] riverside_mike_bifulco_raw-synced-video-cfr_podcast_recording w_0027: Yeah. Got it. Wow. Well, quite the career. You've, you've touched many industries. You know, like, like I guess a lot of successful developers tend to do getting your fingers deep in all sorts of problems and getting your hands dirty with all, all sorts of industries. I guess that brings us to sort of present day and maybe September of 2023, when, when un key sort of raised you round, but also started making the presence online known.

[00:03:26] Why don't you tell us about UN key? What's the story there?

[00:03:28] riverside_james_raw-synced-video-cfr_podcast_recording w_0028: Yeah, so on Key is essentially a. API management platform that we're kind of redefining what API management is and how it's built and how we're using it today. So if you think about API management platforms today, you're talking things like Kong or Tyke or even, you know, Google's offering or AWS is offering through gateways and things like that.

[00:03:50] And when we were building APIs, just in general, my co-founder just built a bunch of APIs. I've done it throughout my career. We found that it was [00:04:00] hard. To protect that API, however that might be. Whether it was through giving people API keys 'cause you want them to be able to interact with it or rate limit that API endpoint for some reason, or it was just a bunch of things that you don't think about until you've already built the API and then it's like, oh, I need to do this and this and this, and there was no easy way to do that.

[00:04:20] It was all or nothing. It was, you need the entire gateway to do a one API key for a user, or you need to, you know, do some sort of ridiculous amount of work to make it globally distributed or whatever it might be. Or you build it in-house, which is where we find some people are still doing. So we kind of had this idea originally to be a blog post for Clark and UPS Dash because my co-founder used to work at UPS Dash.

[00:04:49] We were like, Hey, we could just build a blog post around globally distributed API keys that would like used Clark, and then used up dash. And then basically you link them together and then you give them an API [00:05:00] key and then you can kind of work with it. And after about 10 minutes of talking, I was like, no, this is just a business idea that I think there's actually a business idea that we could build.

[00:05:09] And then the rest is really history. We, for the first commit was in June. The beginning of June, we launched at the end of June, and then everything else has been a rocket ship since then. Lots of users, lots of feedback, lots of different iterations, lots of improvements, lots of new features, you know, all the stuff you kind of do when you're in a small stage startup.

[00:05:31] And then we raised our fund, which was the weirdest thing I've ever done. Worked in all these startups, never seen the other side, and now I understand how that works. Yeah, it was really, really crazy and it was just like you know, let's try us out and see what happens, kind of thing. And here we are today.

[00:05:46] Mike Bifulco: So the rocket ship metaphor is definitely something that sounds apt given the path you've taken to get here. And I am also living on the the, the raising money side of the founder story for the first time. So I can definitely relate to that. If you're listening to this and you haven't.

[00:05:59] Raised [00:06:00] money for, for a project of your own at some point. It definitely changes the way you think about things and requires a pro, a project, a problem, an audience that can hit a certain level of scale. And so hopefully the rocket ship that you've now strapped yourself to is headed in a direction where that scale is really important really important and also

[00:06:15] James: Right. Yep.

[00:06:16] Mike Bifulco: What I'm really curious about there then is like, this is a pretty big problem space. Like building API management is not a, a you know, a small rock in project manager parlance. What was the first cut at this? Like, what, what was the first thing you built to prove that you had something that people would

[00:06:30] James: Yeah, so the first thing, so we decided to go in reverse. So with all these companies, basically doing all or nothing, it was do a gateway and then you can have these features. We decided that the best idea was to give people what they actually wanted, which was just API keys. If you're a small startup, a small business, medium sized business, maybe even a large business, you probably don't need a gateway because you can probably handle most of the infra.

[00:06:55] If you're using serverless, for example, like you can scale your infra pretty well without [00:07:00] having to worry too much about the other features that Gateway brings. So the first iteration was API Keys that had. Per key analytics. So anytime a key was used, you had the analytical data that you needed to figure out, you know, was it a successful request?

[00:07:17] Was it rate limited? Was it use, succeeded? Like a few things, but you could do it per key. So you could really deep, deep dive into your analytics as you were building. And that was the first iteration. So we had rate limiting keys that expired by a certain day. And then we had key, the keys you could attach data to.

[00:07:37] So owner id, metadata was all an option that you could just attach to a key and we would return that data if you needed it. And it was, you could create a key, you could delete a key and you could verify key. And that was it. That was the first iteration. It had a dashboard that had very rudimentary analytical data on it.

[00:07:56] And that, that was basically V zero that we shipped in June. [00:08:00] Then we slowly iterated on things that we thought people needed at the time.

[00:08:05] Mike Bifulco: Yeah. I can imagine just from that first set of features quite a bit of value, right? Possible or at least promise of value possible for folks. From everything from just proving that like, hey, someone is using my API, this person is using my API over and over is a good signal for lots of teams building API products, but also preventing abuse or things like that and, and looking at where costs are being driven up.

[00:08:26] That's super cool. How did you, well, I, I guess you are, someone who's been on the internet for a long time. I'm curious like how you found your first audience for this. How were you

[00:08:34] James: Yeah, so we, both of us are quite known on Twitter for different various things. Like my co-founder is very well known for like open source projects and building a lot of stuff, and I'm very much known for like the YouTube stuff and Clark and a few other things. So when we launched the product, we did a single tweet. A single video, which were spaced out like a day apart. And that really like [00:09:00] gave us the I that actually solidified when we did the launch in June, that solidified the fact that we were gonna raise money because we were getting signals from other founders that we knew in the industry saying like, Hey, this is actually something that people need.

[00:09:15] Because the original project was just a side project. It was just something we wanted to build together. We weren't even charging people at the time. It was all open source. It was just like a fun project for us to build over like a few weeks. And when we did the tweet, it did like 150,000 impressions.

[00:09:33] It brought us like 700 users. Some of those users were just, you know, people checking out the platform. Some of them were actual people that wanted to try the platform. And then, yeah, just a bunch of feedback came through that about like, Hey, this is a cool idea. This is really something we need. No one's really done globally distributed API keys before, like, how does that work?

[00:09:52] And it was very much like, oh, here's the open source project. You can go and look and see how we're building it. And then that brought in some of our first real [00:10:00] users and real contributions at that point.

[00:10:02] Mike Bifulco: I can imagine those things adding up pretty quickly to be the signals you would wanna see. Had you been entertaining the idea of building this into something larger? Were you just hoping to get the open

[00:10:11] James: No. Yeah, we were basically like, cool, this is a cool side project that we can work on together. Andreas and I had never met outside of the internet, like Discord and stuff like that, and we were both very happy with the roles that we had. Like I was super happy at Clark, he was super happy at Up Stash.

[00:10:25] So we used this as more of an opportunity of just like working together, collaborating, building something, having fun. And even when we were being approached by VCs, we have a discord message that we both have the image of, that we occasionally just randomly share in our slack now, which is just us saying like, no, I'm good.

[00:10:44] Part-times fine. Like, I'd rather do this part-time. I'm happy with my job. And yeah, we like, we like to remind ourselves of that occasionally.

[00:10:50] Mike Bifulco: It seems like,

[00:10:51] uh, the universe had different plans for you. Signals like that are hard to ignore. Certainly. And I guess, so at this point, that was last June. This would be, we're we're recording in [00:11:00] early May, I guess you'd call it of 2024. So it's been almost a year. I'm, I'd imagine you've had several chapters of existence since then.

[00:11:07] So tell me, tell me how things sort of fared since then and where you're at

[00:11:10] James: so we're definitely a lot different than where we were. The core product still lies a lot in API keys. Except from now we have a lot more features and we're expanding into more of the API management platform in general. So we can talk about keys first and then we can talk about some of the other stuff we've done.

[00:11:27] So the keys, now we, with the feedback that we've got through customers and just generalized people, we've added other features. For example, now you can do keys with rate limiting that are cost based. So. If you've got like a specific endpoint that you need to rate, limit cost-wise, because maybe it's an expensive resource that maybe it's an AI project and you are giving keys out, you can, you can actually say, Hey, this is gonna cost three instead of the usual one for a rate limit so that you can kind of push [00:12:00] around and, and make sure that like your expensive resources aren't being abused as much as like a cheap resource where you're just looking something up.

[00:12:06] We also have keys that now have. Usage base. So you can just say this key has 100 uses. After that a hundred uses, you can no longer use it anymore. Basically, the only way that it can be reused is if the developer then goes and does an update. So it works really well for token based usage or you know, a monthly subscription that you get a hundred uses of anything like that where you really just wanna limit the amount that someone can use A key, and then we've introduced audit logging and we've introduced better analytics. So like it's more rich in your analytical data. We've added API endpoints so that you can pull the analytical data yourself and then build your drone dashboards or build whatever you need there. And that's kind of how we're fit in, in, into the key space.

[00:12:51] So that's still the core. Real core part of UN key, and it'll always be core regardless if we, you know, once we get into gateways and things like that, it'll still be [00:13:00] core part of our products. And then we introduced standalone rate limiting this year. So if you've ever had like A-T-R-P-C route or something like that, that you just want to rate limit, but you don't really need public API Keys or something like that, you can do that with UN key and it will just be globally distributed.

[00:13:17] So it'll be really, really fast. And you can choose how you want to use it. So we have two options, which is synchronous requests. So that obviously you have to wait for us to actually respond and say, Hey, yeah, they're good, or they're bad, or you can use our asynchronous version, which you just give up a small amount of accuracy.

[00:13:34] So our anchor accuracy is like 98%, even with Async. That will just be like, yep, I, we, we believe this is good. We assume that this is good based upon like the data that we have. So you can have a really, really fast rate limiting system and that's fully globally distributed. It has things like overrides, so if you've got a specific user or maybe an IP address that they need higher limits because maybe you have a deal in place or maybe [00:14:00] you know it's an internal user or whatever it might be, you can manually override.

[00:14:04] In our dashboard and say like, Hey, if this user comes through, they have higher limits. Don't use the usual ones. And you don't have to make any code changes. We'll just say, okay, yep, this person is coming through. Cool. All right, let's give them these limits instead and let's check against those. So basically it gives you this really big amount of flexibility around how you wanna rate limit.

[00:14:25] And those are the big main features that we've added, like since the beginning. We've had lots of iterations behind the scenes of different code base and like we've changed the way that we do keys and what providers we use. And we were originally on servers and now we're all serverless. Like we did a bunch of stuff to like really optimize how fast we could we could get it.

[00:14:44] And yeah, we're really happy with where we're sat right now.

[00:14:49] Mike Bifulco: Certainly, it's a very interesting set of features because you've added a few things that very quickly give any product that takes advantage of it, like a much more polished feel. And [00:15:00] from, from my perspective, they're pretty hard to build. Those are things that you would probably need. If I was going to build that for scratch for an API product, it would take.

[00:15:07] Dozens of hours of engineering time and probably revisits every so often when I un uncover a bug. And, and is the kind of thing that in a world where a AI based products are becoming more and more valuable those features are super important. Like being able to have a customer that gets advanced access or faster access is something that's really massive.

[00:15:25] That's, that's a, a, a great set of features. And I, I like that it's like fairly minimal, but provides, I mean, immeasurable value to, to teams that really need it. That's

[00:15:33] James: Y Yeah. Our whole focus was really around like how as a, as a developer, if I get this product, do I need everything or do I need specific things? Right. And the, the, the problem usually lies that you get everything, even though you don't want it. Then you're like, oh man, I'm paying for stuff I'm not even using.

[00:15:52] Or man, I have to implement this because of the way that it works. We are, we are really focusing on like, Hey, here's a package for rate limiting. Just [00:16:00] use it. Here's a package that gives you the entire API use that, and then issue keys, create keys, do whatever you want. We want people to have the option to pick the features they want versus being forced to use features they don't need.

[00:16:12] Mike Bifulco: It makes a ton of sense and it lets people bite off a, a size of the product that they want to use and that makes sense for their business. And hopefully one of the things I like about businesses like this is that as you are, your customers are successful, you become more successful, you know, as they get more use or more.

[00:16:26] People driving their product and revenue up. Then you're, you're benefiting as well, invariably because they're using more of on key and everyone wins. That's, that's aligning incentives, which is always a good sign too. I'm, I'm curious to hear a little bit about your first customers or maybe your first year of customers.

[00:16:41] Like what are the use cases that stand out to you? How did you find your, find your way into people's hearts as a, a product that they should

[00:16:47] James: Yeah, for sure. I'd love our first ever customer was a crypto trading platform, which is like odd and also kind of interesting. So they were a crypto trading [00:17:00] platform. The name is Premier. They were building an API and we launched pricing right as they were doing their API. So we're like, Hey, we're actually gonna charge for this.

[00:17:09] Here's the pricing. If you know, we'll, we'll happily talk if you're doing big volumes, thinking like, no one's gonna do big volumes on this platform. No one's ever heard of them before kind of thing. And they've been our biggest customer since basically day one. They do millions of requests a month.

[00:17:24] They're pretty popular. They drive a lot of like, resources and we see them, you know, spike when crypto spikes and they dip down when crypto dips down. So we're like, we can see when runs are happening 'cause like our usage will spike and then it will dip back down.

[00:17:39] Mike Bifulco: Sure.

[00:17:39] James: they were an interesting first customer and also just an interesting customer in general.

[00:17:43] Like thinking about how web free is all decentralized and then using a centralized system for your API keys was something that we were still struggling to wrap our head around a bit. But yeah, they've always been, they've been super happy with the platform since basically July. And then a lot of AI [00:18:00] customers coming in where they're building, you know, those AI products that need either API keys or, or something to kind of help them drive forward.

[00:18:09] And it's been interesting to see there that a large majority of our actual customer base that are paying customers are in the AI space. So it's usually ai. And then we have a few Web3 customers, one of them being the crypto trading platform and a couple others. And then we have others like cal.com uses UN key to power their rate limiting behind the scenes.

[00:18:33] Which is, they're, they're a newer, newer customer. They were fairly recent. And then we've just had some side projects where they've come in and been like, I was building an A API and it would've taken me six months to do this work, but with you guys, it took me like a week. And most of that was just like me spinning up API endpoints and then introducing on Key into that.

[00:18:53] And that, that's basically been our customer base. And then people check us out fairly regularly, like two, 300 people a month come in [00:19:00] and check us out and, and play around with the platform and, and see kind of what we're doing. But yeah, that's been most of our customer base. And then we have some other open source friends that also use us for their a their API endpoints too.

[00:19:11] Mike Bifulco: that's something maybe that we've glazed over a little bit here is that you, you have a pretty interesting open source story as well that this is not a, you know, fully private closed source product. So tell me a little bit about that, the, the nature of open source and un key and sort of your

[00:19:24] James: Yeah, so un key is a hundred percent open source. Every part of our platform is there, including our landing pages. The API, there's no trade secrets that are hidden somewhere. Everything is available for someone to go and look at and learn from. The open source nature of un Anki was from day one. We made the decision that if we were gonna build something, we wanted to build it with open source in mind.

[00:19:51] And people are seeing more and more people kind of build on open source, like build successful products. I mean, if you look@cow.com, they're a really [00:20:00] successful platform. And the only thing that isn't open source, I think is just their marketing site. And we just believe that open source makes better products.

[00:20:09] It makes it easier to do security checks, right? Like if you're a big business, you can just go and look through the source code and make sure we're not doing anything nefarious or we're doing something dangerous. It makes it easier for people to collaborate on an idea. So if they have an idea that like, maybe this is a cool feature, they can create an issue and say like, Hey, I'm thinking about this for Anki.

[00:20:30] What do you guys think? And then we have our community. We also have our team. And we can get to a point where an idea may not be the best idea in the beginning, but by the time we've all talked through iterations of it, we've come out with this next feature or an extension of a feature that could really be useful for more than just one person.

[00:20:50] And we, we just found that we just wanted to be as open as possible. So even internally as founders, we are transparent with our team. [00:21:00] So they know exactly what's happening. What are we doing? What kind of deals are in the pipeline, what deals have fallen through or are successful? And then again, like things like revenue and all that kind of stuff and how much money we have in the bank.

[00:21:13] All that stuff we're completely open about. And it just makes for better overall feeling every day knowing that nobody is in the dark and there's no secrets that are being hidden. And we believe that works for software too. Building that way. Makes it really easy for someone to come and learn or, or contribute or whatever it might be.

[00:21:32] Mike Bifulco: Clearly it does. I think you've established as, as a pretty impressive product and one that people are providing value to. Very quickly while we've been chatting, I pulled up your GitHub repo, and I know this isn't exactly a direct litmus, but it looks like you've currently got over 1400 pull requests merged into the product between, you know, now and its inception.

[00:21:50] Which is, you know, from any number of, of contributors you know, between internally and externally. But one of the things I think that comes along with that is you end up having to do almost like crowdsourced product management [00:22:00] too. And, and you touched on that a little bit, but how, how do you balance, like taking feedback from open source contributors versus paying customers versus what your team internally thinks

[00:22:09] James: Yeah, we spend a lot of time talking about this just in general, so like being such a young company, being less than a year old, it it. A lot of our direction comes internally around we believe this is a feature that people want or need because that's something we want or need. Right? Like it's one of those things where you're like, I'm playing around with AI and it, wouldn't it be cool if we had this feature?

[00:22:30] And, and so over the ti over the time period, we've had people ask for feature requests and, and whether that's through open source or being a paying customer. And it's just a fine balance of deciding whether or not it's for the. Greater good of the product, or is it a very niche thing that somebody needs?

[00:22:50] And if it's more niche, we try and pry a bit more into what they're trying to achieve because it may be something we already have on the roadmap, but it's like, you know, [00:23:00] phrase in a different way or, or what they're actually asking for is not what they really want. And we found over time that we, we've, we've really got a fair balance there of like, Hey, this is a cool feature, but.

[00:23:11] Like, here's five different edge cases where this probably wouldn't work properly. And like, what would you think about this instead? And kind of talking to the community in general. And that's been really good for us. I think it's a tough challenge and I think anyone that has an open source project, it's just tough to be like, we are not gonna build this, or we are gonna build this, but it's gonna be six months from now.

[00:23:31] But if you wanna open a feature request and like do it yourself, like here's how you can contribute, and it, it's just tough. I think it's just one of those things, you just gotta balance it the best you can and decide what's better for the greater good of the company and, and the business side of things too.

[00:23:45] Mike Bifulco: By nature you have a lot of constituents through opening the product up, and it's usually a good thing to listen to feedback from, you know, customers and people who maybe just want to use your product aspirationally, but are blocked by, you know, something not existing. But definitely also certainly have to [00:24:00] balance this out as like it needs to run as a business and be a functioning business too.

[00:24:03] There's, there's. Well, almost the curse of open source is that when you're successful, you start to get a massive amount of signal to noise problems. But the benefit of it too is when you strike the balance, like your, your product also a little bit maintains itself or maybe grows itself. And, and your team can, be more open about that, you know, like, let's have a discussion about this feature before we go and implement it. And it sounds like from your ethos of keeping everything, you know, open with the team and, and being sort of approaching the problem with open arms you can ask a few whys, like, okay, you want this very niche feature, but I.

[00:24:33] Why, you know, and why to that and why to that until you get to do something that maybe is a little more generic or you know, I'm sure in, in some cases you're ending up saying no, but you've had the discussion and hopefully publicly too, so everyone can kind of learn from it and chime in as well.

[00:24:46] I'm, I'm I'm, I'm impressed how quickly this has come together. Like it feels like you've got a a timeline that makes no sense to me because it feels like you're traveling through time but also a product that feels really, mature, right. Given that like there's a lot of polish in what you've built and a lot [00:25:00] of really thoughtful touches across your site, across documentation, things like that.

[00:25:03] So I'm, I'm interested a little bit in hearing about the developer experience of the product and from a couple of angles. So first why don't we start with, tell me a little bit about like, the implementation side of this. I, I have a, a AI based product and I want to use un key for you know, managing my API keys.

[00:25:18] How do I get started?

[00:25:19] James: Yeah, so it, we've tried to make on key as easy as possible with the smallest amount of implementation. So if we just use TypeScript, 'cause it's probably the easiest example because we have like official SDKs for that. Essentially the idea is you sign up for an account through, through on key. And we give you an onboarding experience depending on what you want.

[00:25:38] So if you are looking for API keys, we, that's one onboarding. Onboarding experience. If you just want straight rate limiting, we give you a different one. But when you go through that experience, essentially what you get is something called a root key, which is basically the keys to the kingdom. You can set it to be able to do a bunch of different things, but the idea is that root key gives you the ability to create a key.[00:26:00]

[00:26:00] Or update a key or delete a key, whichever one of those kind of operations that you need. So the idea would be that you have an API endpoint somewhere that basically just says, you know, when a customer clicks this button, create a key for them. And it's just a simple, a simple SDK call to like un key client dot.

[00:26:19] Keys create and you give us some sort of information based upon that. So it's just like our API ID and then maybe you need a user id 'cause you wanna reference it to someone. We give you the ability to do that and then we return you two things. The key itself and a key id. And the key is only ever returned one time and we'll never be able to retrieve it.

[00:26:40] We'll never be able to show it to you again. Basically. It's just a one time thing. 'cause we, the way that we store the keys securely. And in that point you give it to the customer and the customer now has this key that whatever it might do to interact with your system, then we just have another piece that's called verify key.

[00:26:57] And Verify key takes two pieces. It's the key [00:27:00] itself and then the API ID that you want to verify against. And then you just make a call to that and we'll tell you whether or not they should have access. So if you've got rate limiting, for example, and maybe someone's really hammering an API endpoint and they've used.

[00:27:14] More than you've said that they can use for rate limiting. We'll actually return back that you sh basically it's a true or false statement and we'll just return false. And then at that point you can handle it however you want. You can reject them, you can tell 'em to calm down whatever you want to do through that.

[00:27:31] And that's basically the way that it works. And then if the user is using something like token, you know, usage based, and they're like, oh, hey, this person's actually upgraded their account. I need to update the key. We have another call that you can update key and then just give us whatever you need to update, whether that's maybe their rate limits are higher now, plus they are, they, they get more usage or whatever it might be.

[00:27:54] And the way that On Key's built, it's built on a really simple rest, API. So if we don't have an [00:28:00] SDK for it, the rest API is really, really easy to use. We have an an open API spec. So if you want to use generator to generate all the endpoints you can. And we've made it as easy as possible to get started.

[00:28:12] In theory, if you wanted to do a next GS app, for example, that had this a like an API endpoint, and then you need to verify the key beforehand, you could probably do the entire UN key implementation in. Less than 20 lines of code, including verifying your keys. It's very, very simple. We've made it as simple as possible and it's really easy to get started.

[00:28:37] Mike Bifulco: I am sure many of our listeners are thinking about the subtleties implied by everything you've just said with a small amount of code. 20 lines of code, you know, for next are probably similar for other platforms. You've just described quite a bit of. Work that is not easy to achieve, right? It's like a very humble description of a super cool product, securely issuing and reissuing keys and mapping them to access [00:29:00] rights and rate limiting and all that is a crazy amount of scope and like a very cool thing to be able to just kind of you know, excuse the slightly sarcastic, joke, but like NPM install and have, you know, your API managed. Like, that's wild. That's, that's such a cool thing. You also touched on something before that I think is really interesting and maybe subtle if, if folks listening haven't touched on this or needed to implement this for themselves, but usage-based API usage when it comes to pricing is a really challenging problem.

[00:29:25] And for the reasons you were saying before synchronous versus asynchronous is a really interesting challenge. So maybe I'll set the scene here and I'm curious maybe you can describe like the problem space and why it's important. But again, let's say I'm building an API product and you've, you've paid me for access to it.

[00:29:39] So you've given me, you know, 10 bucks a worth of API usage, and I have a usage based API call. There are a few things that can happen as you're, you're. The, the value of your credits with me drain having to do with like race conditions and depletion and refunding and all of those sorts of things with my account.

[00:29:56] So how do you convey all of that information to end [00:30:00] users when you're

[00:30:01] James: Yeah, so it is very hard to do if you're not using something like UN key. 'cause usually it's like, oh, I have to use my database to basically store a token number that then I decrement and increment, and then I take the risk of like my DB being slow and now they've burst through and now they're in negative numbers.

[00:30:17] Like how does that all kind of work? So with UN Key, we basically give you the ability to, to put in a number, whatever that number may be. Let's say they pay for a hundred uses for $10. At that point, you have a few options with un key to decide how you want to handle the future of that key. So for example, let's say they decide, oh, this product's terrible and I wanna refund, and you give them a refund.

[00:30:43] They technically still have access to that key when you do the refund. And so what we can do is we have a few different ways to kind of basically disable a key. So you can do it right through the dashboard. So if you're still manually billing and you're manually doing that, you can just go into the dashboard, find the key [00:31:00] that's linked to that user and disable it, and then they won't be able to use that.

[00:31:05] If you are doing automation where, let's say they wanna refund and then you basically scrub their account. You can use either the delete key, if you know for sure. They're never, ever gonna come back permanently, delete the key. And now the, the key has become sort of, you know, unusable or you can update their key and just say it's disabled.

[00:31:24] Maybe they'll come back. And then what we do to protect you from things like race conditions and things like that when you're using our key system. We don't let you do an asynchronous request. We like, you have to wait for us to actually respond. You can't just be like, I'm just gonna assume everything's fine and continue on.

[00:31:41] And that protects you because one, it basically stops your. Anything else being processed until UN key responds. And because of the way that we work behind the scenes, essentially what you get is the correct and valid number. Like we are heavily cashed. We know exactly what the number is at the time when the [00:32:00] user makes the request.

[00:32:00] So there's less of this worry of like, I have to wait for my DB to respond. Oh, they've burst through. There's no way for them to actually physically burst through that request. And then what you can do is like once they deplete and get to zero. Maybe they wanna pay you $10 more and say like, oh, I want another a hundred users.

[00:32:16] This is super cool product. Well, you can do the same thing. So you can either manually just do an update call and say, give, update remaining, and say, give me a hundred more. Or if it's like a monthly subscription, we have the ability to refill automatically once a month. So every month, if they've billed, you can just basically say, every month, just unie, handle this for me.

[00:32:38] I don't care about it. Just handle it. And that makes it really, really easy. And then the final piece to the puzzle is in the very near future we, we will have web hooks for you to hook into. So we can tell you this user has done this much usage if you're doing usage based billing on your side, and we can send that to [00:33:00] you in specific timeframes or usage base or however you wanna really do it.

[00:33:05] And so that you can bill via like 30 day invoices. So like, let's say it's similar to us, where once a month you get a bill from us and we tell you how much you owe us. You can now do that with UN Key and not have to worry about implementing the entire thing on your stack and use whatever billing partner you want, Stripe, paddle, lemon squeezy, whatever it might be.

[00:33:23] Mike Bifulco: I suppose that means somewhere under the covers you're probably using UN key to

[00:33:26] James: Oh yeah. Un key is built on un key. That is for sure. Yeah, we, we everything that we do is powered by un key. It's always the classic trope of like, you know, Clark is using Clark behind the scenes, right? Like, if you go to Clark's dashboard, you have to log in with Clark to get into Clark. It's the same thing for us.

[00:33:42] So when you. Bill when you like create your workspace and you create your API and you create that first root key, that root key is then using UN key to actually build you a root key that you then use for a specific account. It's very similar to that. And then like when you do a verification, we verify that key against one of our [00:34:00] own keys.

[00:34:01] Mike Bifulco: That creates, a paradox of its API keys all the way down, but that's, that's a great way to build the product, making sure that you're close to it and using it. I I also wanted to talk a little bit about maybe the other side of, of your product in DevX. And particularly with your background in developer advocacy it feels like you've done a really good job of, you know, making a splash and socializing this because of the reputation you built for yourself.

[00:34:21] But I'm curious what what other I. Benefits your work in developer advocacy has, has brought to engineering a product and building something that actually used. And maybe the additional caveat that I'll add there is that I, having worked in developer experience for, for a while myself, one of the things you hear, especially while interviewing is that developer advocates aren't engineers.

[00:34:39] Like, don't build products, can't build products. And, and I think that. Obviously is a, a trope and something that has a lot of faults to it, but it's something that also just doesn't get talked about, right? Like you have jumped from the world of talking about building for developers to building for developers, and now you've got a product that's out there.

[00:34:54] So yeah, tell me your secrets, man. I wanna hear it all.

[00:34:56] James: I, I think, yeah, that classic trope of like. [00:35:00] Dev advocacy never can build a product. Is is, is something that I think was probably true when dev advocacy was new, right? Like it was just like more of an educational kind of part than it was about anything else. And less about being involved in like moving a product forward or pushing a product forward, or helping engineering or whatever, helping customers, and, and because my development background was way before I was dev advocacy, they gave me some of that skillset immediately, right? Like I could build a product and don't have to worry. But I think having the background in dev advocacy and being unafraid to be open is something that's scary for a lot of founders, right?

[00:35:45] Like you see these cool products like one time and then you never see or hear from them again. And then like two years later you find out, oh, they, they shut the doors 'cause they couldn't get another round of funding or they just couldn't get traction in the space that they're in. And it's because you never [00:36:00] see them actually doing anything on either social media or blog posts or being on podcasts because they're afraid that whatever they put out in the world. That's the be all and end all. Like it's gonna be cemented in history and I can never go back on what I've said or I can never change my mind or I can never improve the products because we've said this thing. We took the opposite approach was just like, just even if we're in the middle of building sand, just post it on social media and say like, Hey, this is coming soon.

[00:36:28] Or Hey, we built this and it's like in alpha, like, love some feedback. Then doing podcasts and, and, and blog posts about things that we have opinions on is a huge thing. 'cause you know, the internet is full of people's opinions and having our own as we've built this product. Like we have a blog post that's about you UIDs and how like people will use them wrong and like, this is how you should really try and use them and here's why.

[00:36:52] And that post is still one of our biggest drivers to the product. And most of it comes from a [00:37:00] place of like, either, wow, this blog post is amazing, it's a really great read, and now I wanna know more about products. Or, these guys have no idea what they're talking about. They have no experience. Like, they're clearly wrong and like, I have a really bad opinion about this.

[00:37:14] Right? So like you have these two opposing opposite ideas and, and, and we just don't care. We're just like, cool. Thanks for coming and checking us out. Like. That's your opinion and you know, you rightfully can have that opinion. And building all those things together kind of really helps kind of cement the products as a whole.

[00:37:34] And then having this engineering background and, and not being afraid to like be wrong about something. I am nowhere close. My ability is nowhere close to my CTO's ability in engineering. So there's times where I'm just like, ah, I just put this together and see what happens. And I push PR up and he is like, what are you doing?

[00:37:53] Like, now this makes sense? And like, okay, well let me go and try again. Like, kind of thing. And not being afraid to, to, to [00:38:00] learn as we go. Because a lot of the stuff that we built, we had no idea what we were doing in the beginning. We were like, we think this is gonna work, but we won't know until we try. And, and just being unafraid of that.

[00:38:11] And I think dev advocates are the perfect people for that too. A lot of the time, you spend a lot of time building content or education or talking to people or whatever it might be, and there's a big chance that whatever you put out in the world, either people aren't gonna like, they hate the, like they really love it, it's not very good, or it miss them up completely and being able to like get back up and be like, oh well, like we did our best with that one.

[00:38:38] Let's move on to another post or another video, whatever it might be. That's the same I attitude that, that you need to have, if you're gonna build a product, it's just like, this might not work, but let's give it a shot and see what happens.

[00:38:51] Mike Bifulco: I love that, that's like a masterclass in both developer advocacy and startup founder school, you know, kind of slammed together. There is I feel like [00:39:00] the, the, one of the, the hidden superpowers of developer advocacy is learning to embrace opening and working in the open and in public and, and benefiting from the.

[00:39:09] Internet's magical superpower of every time you share something online, someone smarter comes along and tells you why you're wrong. That's fantastic. Like as, as long as you look at that as, Hey, I can learn from this. That's, that's a superpower, that's a skill for sure. And I think a lot of founders, I, I mentor a handful of startup founders and kind of give advice to folks who are like early on in the journey.

[00:39:26] And one of the things that I think people miss out on a lot. That a lot of building a product, especially a developer focused product, is that you've gotta love building it and rebuilding it and iterating and like every step along the way is part of the thing. It can't just be like, I did it once. It's done.

[00:39:39] It's never done. And that's, that's the fun part, right? You've gotta learn to love that too. It seems like you've married the two together really well, and certainly your team has too. We actually haven't really talked about your team much apart from yourself and your, your co-founder. But how big is the team right now?

[00:39:51] What, what are what's the scale you're at?

[00:39:53] James: five people total, including myself and Andreas. We have two engineers and then one design [00:40:00] engineer who does both design and also a lot of front end work for us.

[00:40:05] Mike Bifulco: Yeah. Got it. Yeah. A small but mighty team building something that's really, really pretty wild. Are you expanding at the moment? Are

[00:40:11] James: are not hiring right now. We just hired just recently, actually April tonight, design engineer came on board. So we'll probably be hiring again core, kind of close to the end of the year. We definitely don't wanna expand too far, too early and then be like, well, well I don't know what you can do right now.

[00:40:29] 'cause we're like, we're all doing these tiny pieces, you know, or the, you know, the, the scary worry of being a founder is like, what if we run out of money? I don't wanna have to tell you. Sorry, like you just got a job with us now we have to let you go. But yeah, we'll definitely be hiring kind of close to the end of the year again which is when we started our first hire round.

[00:40:48] So probably like October, November, we'll probably start hiring again.

[00:40:52] Mike Bifulco: Yeah, that's great. I will make sure by the way, to include links to UN Key and to James and to all of the social, everythings in the show notes, [00:41:00] as well as actually the blog post you mentioned before. Before I let you go, I'm curious to hear what's next, like, what are you thinking about, what's the next exciting feature

[00:41:06] James: Yeah, so we have a couple of different things. So we're currently working on a special project for AI users. Essentially the idea is semantic caching for ai. So the idea is that we can save you both money and also just speed and latency by using the same tech that we use behind the scenes but give you the ability to, to really have some really nice responses if it's something that's already been asked before and you don't have to go and ask open AI the same thing or some variation of that.

[00:41:37] We're working on that right now. That should be released probably in June. If all goes to plan. And that will be our first real, like push into kind of the gateway esque product. And then the really big thing in the next probably 12 months will be gateways in some variation, but gateways that developers can use instead of gateways that you need a whole DevOps team for.

[00:41:59] We wanna make [00:42:00] it really easy. A few clicks, gateway, or here's an API call and now I have a gateway. That's the idea. We're pushing towards the same devex experience that you get with the rest of the UN key products. And then we just have a few more things coming. Like we have a cash package coming that's not really related to On Key, but is where we can help you do a lot of like the heavy cash work in CloudFlare workers and things like that that we use.

[00:42:23] Now we're actually building that out to be just a standalone package that you can use for whatever you want and don't have to use on key. And then, yeah, it's just some generalized improvements, more analytical data for you guys to ingest and play around with. Ingestible audit logs is coming so you can actually push stuff to on key and then do something with it afterwards.

[00:42:42] Lots of things just to try and help API development in general

[00:42:46] Mike Bifulco: There's never any shortage of new features to build, especially when you've got an audience of

[00:42:50] James: Yep. There's never, there's never a time where we're like, Hmm, what should we build now? We definitely know that there's things to build and, and we're happy to build them.

[00:42:58] Mike Bifulco: james, that's a really good spot to be in. I [00:43:00] appreciate you coming to hang out. Last thing where's the best place to find you online and what's the address to find UN

[00:43:05] James: best place to find me is probably Twitter. It's where I spend most of my time. It's just James R. Perkins. You can find me there. And then for Un Key, you can now find us@unkey.com. We just launched the new site that you see today, which will be whatever anybody looks at right now. We launched that a couple of weeks ago with un key.com domain.

[00:43:23] Mike Bifulco: it looks great. Congratulations on the launch James Perkins. I am thrilled to be able to talk to you and super excited to hear about your story. Please come back anytime when you've got launches to chat through or things you wanna ask the community, or if you're just interested in getting yelled at about open API specs, I can do that too.

[00:43:38] Thanks so much for joining me today. I really

[00:43:39] James: Thanks, Mike. Really appreciate you. Bye.

[00:43:43] Mike Bifulco: Soon.