Algora | Algora TV

Open source interviews #30 - James Perkins, founder of Unkey.dev

Algora

Watching now

00:00 Hey everyone, welcome to the Open Source Founder Podcast.

00:04 Joining me today is James Perkins, co-founder and CEO of Ankit.dev.

00:09 James, thanks so much for joining. I'm very excited for this.

00:12 And we'd love to start with some information about your personal background

00:15 all the way to how Ankit got started. You met Andres.

00:19 Yeah, for sure. So my name is James Perkins.

00:23 I've been in the industry for the last 15 or so years.

00:27 I started my career at a startup that was basically funded by banks. So it wasn't VC funded. It was funded directly from a bank.

00:36 And then I've moved around my career. I've been in Fortune 500 companies all the way back down to startups again um my most recent startups before unkey i worked

00:46 at plaid which is a fintech company i worked at tina cms which is an open source uh cms that

00:53 basically uses mdx plus git and gives you like visual editing and then my most recent one before

01:00 i left to go to unkey was clerk which is an authentication provider, which is quite

01:05 popular now for React developers. And then yeah, then Unki came about. And we can talk about how

01:13 that came about, because the story is kind of interesting. I met Andreas probably a year before

01:22 Unki was really started.

01:27 We had interacted through a couple of different channels.

01:32 First, I helped Andreas when he was working on Planetfall,

01:36 which was very, very early on in the app router world.

01:38 It was like next year's 13, you just come out.

01:42 And typical Andreas, bleeding edge,

01:44 right on the very edge of bleeding,

01:45 was working on that. And Clerk had released a package that did support typical Andreas bleeding edge right on the very edge of bleeding, um,

01:45 was,

01:46 was working on that.

01:47 Um, and clerk had released a package that did support app router,

01:50 just had some rough edges because we were kind of working on a very short

01:54 timeline.

01:55 Um,

01:55 and so I helped Andreas get through that.

01:57 And then we stayed in touch throughout that period of time.

02:01 I did some work for upstash,

02:03 um,

02:04 some consulting work and, and some advice and some

02:06 stuff around DevRel and marketing and all those fun things that are really hard to get started

02:11 when you're in pre-seed or seed. And then Andreas pitched me one day, hey, I have this idea

02:17 for co-marketing, which was essentially how Unki started. He wanted to basically take Clerk and Upstash

02:27 and kind of make sort of what Unkey is today,

02:30 which was basically like,

02:31 how could we do this in a way that used Upstash

02:33 and Clerk together

02:35 and make some sort of API key management system,

02:38 make it really easy for people

02:39 to basically handle some of this stuff.

02:42 And then he pitched it to me and i said no let's let's do

02:47 this instead like we talked about it for a little bit internally over discord and i was like we

02:51 should just do this um we hadn't really worked on a project together uh before that um and then we

02:58 just took the idea and said let's do it and see what happens uh so it took us about seven or so days

03:05 from idea to like an mvp-ish and that was like a lot of copying and pasting code from other

03:12 projects that we'd done over the years and it just kind of slapped it all together

03:15 used templates from like tailwind and all sorts of things like that just to get the idea out there

03:20 um and then we launched it and apparently people wanted it um it got very it did very well

03:28 like the first week we we did like 750 stars uh we had almost a thousand signups um and then

03:36 within july we were like okay well this seems viable we should add some sort of pricing

03:41 so that we can potentially monetize this or at least find out if

03:46 people are even interested in paying for a service that does this idea being 100 open source and

03:52 already we were kind of like maybe people don't want to pay for this and maybe they'll

03:56 want to use it for free but maybe there's no monetization there um and we monetized it and

04:02 and a few days later we had our first paying customer.

04:07 And they weren't a small customer.

04:10 So it wasn't just like a random developer.

04:15 It was a fairly large customer at the time who was building out an API.

04:16 And it just happened to all kind of align.

04:21 And we had all the features they needed at the time.

04:25 And so we talked to them and we were very excited.

04:26 And then we were like,

04:28 well, if this customer is coming on,

04:29 we're probably going to have to incorporate because neither one of us want the tax burden

04:32 of the payments that come from this.

04:35 So we incorporated.

04:36 And as we were incorporating,

04:38 we got introduced to a bunch of VCs.

04:42 And then that's kind of how the,

04:44 hey, maybe we could get this funded discussion started

04:48 um and before that we we there is a famous discord message between andreas and i that says like

04:56 we are both very happy with our jobs let's keep doing this part-time let's not try and get funded

05:02 let's just do this as a side project and occasionally

05:06 yeah occasionally it ends up in in our slack messages it's like remember this um as a as a joke

05:14 um but yeah we which then started the vc thing which i can happily talk about um and and how

05:22 fundraising kind of works and and kind of how that uh progresses and especially

05:27 in open source it's a bit of an interesting conversation um but yeah that's kind of how

05:32 unki gets started and now we are here uh you know almost six months later i love it uh it's phenomenal

05:39 thanks so much for sharing the story so far and it sounds like an overnight success shipped nvp in a

05:44 week and then got paying customers

05:46 even before being incorporated.

05:47 Of course, there's this long career

05:49 both of you guys behind to lead to this.

05:52 And for people that don't know,

05:54 Andreas, he's Cronark on Twitter.

05:56 That's maybe how a lot of people see him.

05:58 And have you guys met in person so far?

06:00 Nope.

06:01 We've only ever talked through a camera.

06:04 Nope. Nope. We've only ever talked through a camera.

06:13 We are going to do an Anki meetup in sort of March, April time, where we'll get the whole team together.

06:16 We can actually like see each other without seeing each other.

06:21 Basically, what you see today, which is just like from the waist up, is pretty much all we ever see of each other.

06:25 That's great. And yes, speaking about the team uh how have you expanded uh the group uh up to this day yeah so we after we got funding um i was the first person

06:34 to actually go full-time and that was pre-funding so it was about uh we we'd done the funding deal

06:40 we knew the money was there we just didn't actually have the money um so i i

06:45 started in september um andrea started november 1st i think it was november 1st because he being

06:53 he's so for people he's in germany uh which means there are german laws around how you can leave a

06:58 job so depending on how long you've been in that job you have to give a notice based upon that so

07:02 it can be anywhere from 30 days all the way up to some ridiculous amount that's very very long in the future um so we had to kind

07:09 of play across that um and when we did that uh we got the funding and i said to andreas like hey i

07:18 know a junior dev that doesn't have a lot of experience. But I think, you know, I've known him for almost 15 years myself.

07:28 I think he could actually do the job and do it really well.

07:33 And I would bet my shares of the company on it.

07:36 Okay.

07:37 And Andreas was like, okay, let's meet and let's talk.

07:40 And then gut feeling, gut check, let's go from there.

07:43 And so he met with him, was like, yeah, I think it will be okay.

07:47 So we ended up hiring Mike, who's been our junior dev,

07:51 and he's been crushing it since he started in late October.

07:55 And then January 1st, technically January 2nd,

08:00 our second team member started, another developer named Dom,

08:04 who came from Vercel, has like an engineering slash sales engineering background.

08:09 And so yeah, it's super exciting to kind of make the team grow.

08:15 And it's, you know, an experience to go and hire people and think about like, what part of the team do they bring versus like being part of small departments?

08:22 Because I've been a manager for in

08:25 multiple places and you only have to really worry about your department and that's the only thing

08:29 you have to worry about but when you start a company you have to worry about like how do they

08:32 fit into the company and what do they bring and you know even if they have these weaknesses how

08:38 do the strengths of whatever they have kind of kind of play out um so that was an interesting

08:42 experience for both uh andreas and i

08:45 and then we have to learn about you know how do you pay these people and like all those things

08:49 that you don't really think about when you're just an employee you're just like cool i hired this

08:53 person and they get paid then you have to worry about like how do they get paid and how do you

08:56 stay compliant with each country that you pay people in all those kinds of things i love it

09:01 this sounds excellent our team uh assembled and so everyone's excited for what's to come.

09:07 As a matter of fact, if you could share right now,

09:10 is there something that people can expect when it comes to a next big release

09:14 or something you're cooking?

09:17 Yeah, we're cooking quite a lot.

09:18 But the next thing that is really on the horizon for us

09:23 is moving into authorization. right now unkey is

09:27 only really truly authentication so you get a key you verify it if it says it's true that's it

09:33 there's no uh way to say like add a role or add specifics to say like this api key can only be a

09:40 read and this one is write and delete like you can't do any of that right now.

09:47 And that's what we're working on right now.

09:49 So we're going to have role-based authentication.

09:52 So our back pretty soon.

09:57 And then we just released today our CLI auth examples.

10:00 So now you can see how to do CLI auth from end to end.

10:03 So you don't have to manipulate JWTs or anything. You can just use unki to essentially give someone

10:05 authentication through CLI.

10:08 Exactly. And for that, I think you used

10:10 clack. You made a YouTube video about it too.

10:12 Yeah, so we didn't. I don't think

10:14 we used clack for this

10:16 example. This is

10:18 another CLI that I have cooking in the background

10:20 on how you can contribute to unki

10:22 without having to spend

10:24 so much time bootstrapping it

10:26 essentially it'll be like two commands and it will download that the github repository put

10:31 everything in the right places and then you don't have to worry uh but i use clack for that which uh

10:36 is is really really nice if you know typescript javascript it makes clis really easy so speaking

10:41 of the youtube video actually i think uh well personally i would

10:45 love to hear about your youtube career so far and then of course how it has also helped today with

10:51 you know distribution uh and marketing for uh anki uh followed by you know hopefully advice

10:58 to other content creators who or people who want to get started, uh, super curious to hear your advice here. Yeah. So the YouTube career, uh, I started during COVID. Uh, it was just one of those

11:10 things where I was like, well, I have all this free time now. I'm not going anywhere. I'm not

11:13 doing anything. Uh, what could I do to, to, to maybe educate some people I've been in the industry

11:19 for free, you know, like 12 years or so. And I'd done a variety of jobs. So I thought, well, if I know this stuff,

11:26 like maybe I can educate somebody on pieces of content

11:30 or how to learn or how to become a developer.

11:33 All these kinds of things kind of popped into my head.

11:35 So I started with what I knew at the time,

11:38 which was for a long time in career,

11:40 I was Java developer at enterprise level.

11:43 So I did a lot of like pre like intros

11:47 into java and like how that works and and and how you can manipulate data with java um and and on top

11:55 of that i was doing things like java questions that you'd get in an interview so those kinds

12:01 of things that like are super helpful when you're trying to get a job um and

12:05 over time i realized that that was really boring uh for me not for the for the content people that

12:11 consuming it was great right but for me it was so boring there's no challenge to it it was not very

12:16 exciting um which then led me to to kind of move from that into what i was doing daily, which is like React development, but building it at a level

12:25 of crash course level. So my initial YouTube career started as Java, and then I moved into

12:33 web dev and it started as crash courses. So these videos were anywhere from 40 to like three hours

12:39 long. And that really kind of helped me, one, understand how people learn and two, how to

12:46 market myself as a person, uh, which I think now plays really, really well into Anki, which

12:52 is like, how do you market a product that nobody's heard of, uh, in a similar vein,

12:56 when you start YouTube or you start a blog or you start TikTok or whatever your social

13:00 media platform of choice may be today.

13:04 Um, nobody knows who you are.

13:06 And there has to be some sort of excitement to actually care.

13:10 And that goes through either producing really high quality content

13:14 and then being also really enjoyable to watch.

13:19 And usually what happens, at least in our space,

13:21 it's kind of changing recently.

13:23 Like if you look at who's popular in space, if you look at like Theo or Primogen or any of those kinds of people,

13:30 they found the way to educate you while also being incredibly entertaining enough for you to want to stick around and watch the next video.

13:38 And some of the older YouTubers that have been doing it for years, like Brad Traversey, who does crash courses and has done those for, I don't know how many years now. I remember watching videos and giving them to my

13:49 new employees or juniors and be like, here's how you learn this. Just you watch Brad. It's like

13:53 three hours long. Enjoy. It's still doing that same content and it still does really, really well.

13:59 So I think it's a fine balance of being entertaining and educating and just making

14:03 sure that the content you put out is high quality um and if i was to start today my advice would be for any new person

14:12 is one pick something that you actually enjoy don't think about how do i make this into a you

14:19 know a side gig where i get paid loads of money like Just throw that out of the window and think about, if I did this for 10 years, could I do it every week? And if I can't do this for 10 years,

14:30 and yeah, sure, it'll evolve. But if I can't do the content I make on day one in 10 years,

14:35 then you probably are picking the wrong topic. You're probably thinking,

14:39 this is going to be really popular, so I'm going to make loads of money off of it and then you're just going to get burnt out and bored um pick what you enjoy um and then just start it doesn't have to be the

14:51 highest quality you don't need a 400 microphone and a 500 mixer and a thousand dollar camera

14:57 start with your macbook uh if you've got a macbook just start with that camera and and and any

15:04 microphone will do even the macbook microphone is pretty good today and you can buy lav mics for $20.

15:09 Um, and then if you're making blogs, just put it out there, just start putting the content

15:15 out there, right.

15:16 Daily or weekly or whatever cadence works for you.

15:20 Um, and then that kind of shifts into, to how do you market a company it's the same idea

15:26 um i wrote this really long blog post that broke our own blog um which was about how we market our

15:33 unkey um and this was kind of a transitional thing so when i wrote this blog post originally

15:39 some of it was from how do we do devrel at clerk. I wrote this really long piece about how DevRel was basically dead today

15:46 and the way that people were doing it were wrong

15:48 and the old traditional ways of DevRel doesn't work,

15:52 which if you haven't been around long enough,

15:54 maybe you've never spent a lot of time with DevRel.

15:58 The old way of DevRel was like code tutorials

16:01 of how to get started with a product.

16:03 And that was basically it,

16:04 but it was just different layers of the same thing. Really simple, really shrunk down to basically nothing.

16:10 So the value that a user gets out of it is only at the scale of, if I've never seen your product

16:16 before, this is how I'm going to get value from it. Or I've never interacted with this product

16:22 before, this is how I get value. And when I went to Clark, I was basically DevRel.

16:28 And my attitude was, nobody cares about getting started.

16:32 They can just go to the docs, read the quick start guide.

16:34 It's the same value.

16:36 So let's just do content that makes sense.

16:39 So the first thing that we did was we did a lot of large form tutorials

16:45 that actually solved an issue today.

16:49 Whether that was like, how do I introduce RLS plus Clark

16:53 to make like authenticated with super base?

16:56 Like things like that, that really make sense.

16:59 And people actually need that to solve a problem,

17:02 whether they're building an app, a side project, or a SaaS.

17:05 You need that piece to connect everything together.

17:09 And so we started there.

17:11 And then that kind of grew into, well, I can only do so much.

17:17 I'm one guy.

17:18 I can only put out X pieces of content a week, a month, a year.

17:22 And I can only talk to 200 podcasts.

17:25 And then people are going to get bored of me, right?

17:27 If they hear my voice all the time,

17:28 you're not going to listen.

17:29 So I went with this strategy of finding people

17:33 who I believed once they tried the product,

17:37 they would be a super fan essentially.

17:41 They would love the product so much

17:43 that I wouldn't even have to approach them

17:45 to say, hey, if I want this piece of content and you're a popular blogger, YouTuber, TikTok,

17:52 Twitch, whatever, if I approach you and said, could I sponsor you? The odds of you are answering

17:58 with a yes is very likely. And then I know whatever you put out is going to be genuine enough to make sense.

18:07 So I scouted the internet looking for people, friends of mine,

18:12 people that I'd interacted with over the last few years,

18:16 and I came up with this list and I put it in an ocean doc and said,

18:20 okay, these are the people I want to try club one time.

18:23 And once they try it one time and once they try it one time

18:25 if they hate it that's fine just remove them from the list if they love it and it suddenly

18:29 solved some sort of issue for them then we can talk about how can we sponsor this person

18:35 um and so for clerk it was theo so if people know theo theo is very popular in the space now

18:42 um and and even back then i think he only had

18:46 so now he has 200k subs i think back then it was at 50k or something like that it was

18:51 a lot less than he has today but he had an audience that loved what he was doing

18:55 um and i badgered theo for months and months about just random things like we were interacting with each other we were you

19:06 know doing all that kind of stuff and then i had realized talking through him he has create t3 app

19:13 which is popular open source project and they had a create t3 turbo app which is also a popular

19:18 open source project which is essentially t3 on both exppo and Next.js.

19:29 And the problem they had was they could always authenticate the Next.js app,

19:31 but there was no way for them to produce some sort of good authentication for the Expo app and have them interact so you could use one TRPC.

19:36 So I said, okay, I can build that for you with Clark, and it will just work.

19:40 So I built this repo out.

19:41 It's still there at Clark.

19:43 I'm sure they're still maintaining it

19:45 but there's there's a there's a project there and i said okay here's the solution for you

19:48 let's make a video and so he tried clerk a couple times he was like okay yeah let's do the

19:54 sponsorship so we did the sponsorship and uh the first video ever he did he didn't even show any

19:58 code all it was a three minute video and we intentionally showed no code at all we just

20:05 showed the app working from end to end and showed how it worked and then like put a link to the

20:11 github and said like this is how you can do it and then the week afterwards i did a video on my own

20:15 channel that described like how the actual pieces fit together so now we've made two pieces of

20:20 content and i've paid theo once and then theo was like oh cool awesome a video let

20:25 me tweet about it let me post it on my own youtube as a community post and then that evolved into

20:31 okay well i'll make a blog post now so that can be found on the internet and so now we have like

20:35 three or four or five different pieces of media and then that's how the relationship built and

20:41 then we went further and further and theo and i are now friends. Theo actually invested in Unki. And that's kind of how you have to start your trend with a new

20:52 startup. And it takes time. In the open source world, it's a bit easier, probably, because

20:58 people that contribute, you can probably figure out, how can I make this person who contributes?

21:03 Maybe they write blog posts or something like that. you can kind of angle it um but it's all about finding those people and once

21:09 you find those people um you can go from there and then things like twitter and stuff like if

21:15 you have a business account for your twitter account just don't use it uh is essentially my

21:19 opinion and people get pretty upset about that but But you're better off making your own personal account

21:25 and using that personal account to promote your business.

21:29 If you look at Unki,

21:31 I think we post retweets and quote tweets

21:36 and that's about it.

21:39 We've never really posted anything on there

21:40 because people trust someone with a face,

21:43 but they don't trust the business.

21:44 I'm not going to follow a business. I'm not going to go and be like, Amazon, AWS, let me follow the

21:50 AWS account. I'm going to be like, who's the DevRel at AWS? Okay, it's this person. Let me

21:54 follow them because they're going to have nuggets of information that I care about.

21:58 AWS is probably going to try and promote some new thing that I don't want. I'm just not interested

22:04 in it. And then as you build your

22:07 brand up you can then start using your own twitter account or your ex account whatever we want to

22:10 call it today uh to do things like change logs or um things like that but you got to keep it fun for

22:18 someone to interact with so if you look at my twitter and you look at Andreas's Twitter and you look at our Unky's Twitter, it's very much a lot of shit posts.

22:28 Because one, it's social media and it should be fun for everybody.

22:32 And two, people like that stuff.

22:35 They don't want to come to Twitter and be educated.

22:37 They want to have fun.

22:38 And maybe there's some education going on, but it's like very lighthearted.

22:42 And doing those kinds of things like lets people know that

22:45 you're a real human you're not just like some guy some millionaire with a ferrari like i'm not any

22:51 of that i have a car that's like 20 years old and i live in a house it's pretty small like i'm not

22:56 some millionaire i'm just some regular guy who just happened to build a company and um i think

23:00 that gets lost as as people try and build their business up.

23:06 Well, okay.

23:08 That was phenomenal.

23:10 Thanks so much for taking the time to say all this.

23:10 Sure. In fact, the story with Theo was excellent, backing what you said earlier.

23:14 And, you know, I can second the fact of how anyone can just get started,

23:19 even if you don't have the gear and the setup and everything.

23:22 I mean, when starting this podcast last year, I did the first 25 interviews just recording a Zoom call.

23:28 I didn't even with the max, you know, microphone and camera.

23:31 I know it didn't look great, but I went through those reps.

23:34 And so it was super useful.

23:36 And now trying to step it up a little bit with this new season.

23:40 Thanks again for joining.

23:42 And that was excellent.

23:44 That was a lot of great advice.

23:45 That part specifically about on Twitter, for example,

23:50 and on all channels utilizing your personal account

23:52 as opposed to the company account.

23:54 I mean, when we had this back and forth over the edge of Discord,

23:57 and you told me that, that's when I also started my personal account,

24:02 still not a YouTube channel.

24:03 It's tough when you start from zero for both,

24:06 but your advice makes sense here.

24:09 So thank you.

24:10 Thank you so much for that.

24:11 Going back to the business end, if we can,

24:16 you already discussed distribution basically,

24:19 and, you know, with all these activities that you do,

24:22 you said you expanded your team to go through the standard activities

24:26 for a dev tools company,

24:28 software engineering,

24:29 and then distribution,

24:30 dev rel, marketing,

24:31 and then you have customer success.

24:32 And probably later,

24:34 eventually you have some sort of

24:35 enterprise sales motion

24:36 or any outbound motion.

24:37 Is that something that as a CEO,

24:39 you today sort of like map out for the future?

24:42 Are you going to let it come at its own pace

24:44 and then you know scale

24:46 the team accordingly how do you how do you approach this aspect yeah so so one being a ceo is is is

24:54 kind of a strange concept um never done it before so i'm still kind of learning ropes but i have a

25:02 lot of founder friends so that that makes it a lot easier, right?

25:06 But the key aspect here is we're taking an approach

25:12 that for some people may seem,

25:15 oh, they're doing it too slow

25:17 or they're not making risks for growth, right?

25:23 So the way that we've kind of handled this is I'm not willing to scale the team if we can't do it ourselves.

25:32 So for marketing and distribution, all those things, I'm doing all of that right now.

25:36 So I do all of the change logs.

25:38 I do blogs.

25:39 I do basically anything that's marketing-based, I do.

25:43 And then moving forward with that,

25:48 Andreas takes care of the tech side of things.

25:50 He takes care of the team for the most part.

25:52 And I do all the paperwork that comes along with being a business.

25:55 So I make sure the accounting and the bookkeeping,

25:57 all that stuff is done and done correctly.

26:00 And our accounting team have whatever they need.

26:02 And then as that goes on, so enterprise and outreach and things like that,

26:10 we're doing that ourselves right now so right now we're still leveraging friends is not the right word we're leveraging contacts that we have in the industry right so

26:18 we know a lot of people that are in our space based upon a couple of things whether it's like

26:23 we know they have an api and maybe they're not doing it right.

26:27 Um,

26:27 or maybe they're thinking about having an API and,

26:31 and we've talked to them in the past and maybe now's the time to start

26:34 talking to them again.

26:35 Um,

26:36 and then on top of that,

26:37 because of the word of mouth and things like that,

26:41 it makes it very easy for these business plus slash enterprise customers to actually come to us and reach out to us.

26:50 We probably won't hire a salesperson for a long time just because one, enterprise is not our actual target audience today.

27:06 audience today we're more targeted on startups and and business medium business size customers who need these features today and don't already have like an ingrained system um and sure in the

27:13 future we'll probably start tackling those people and scaling the team at that point will be a

27:18 decision that i probably make and then andres will you know we're 50 50 partners so i'll talk

27:24 to him about like hey hey, I think we need

27:26 this type of person. And we'll talk about it internally and we'll make a decision. And then

27:31 once we've decided if we think it's a good idea, we'll actually pitch it to our own team.

27:36 So most decisions that happen at Unkey is a complete... Everybody at the company makes

27:40 these decisions. For example, we just started sponsoring open source software.

27:45 So we did Drizzle just yesterday because we love the Drizzle team.

27:51 And then they do really good work.

27:53 So I basically went to Andreas and said,

27:55 hey, we should give some of our VC money away

27:58 by sponsoring people that we're using.

28:00 So we did Drizzle and then we said to the team like,

28:03 hey, we're going to allocate $500

28:05 a month. We've given this much to the Drizzle team. So we have this left. Who else or what else

28:13 in our stack would you like to sponsor? And then we'll filter those out and then we'll make that

28:19 decision. And similarly with Mike, Mike was already hired before Dom was. So we gave Mike the opportunity to actually sit down and talk to Dom and be part of the entire hire process because we wanted him to feel that I'm part of this team and yet sure I have equity in the company, but I'm actually making the decisions that drive the company forward and sure there's some decisions that i just make and like the ceo makes those decisions but for the most part we we try and make it a collaborative effort

28:48 because just like open source like our company is very open and and we're open about everything

28:54 with our team financials where we're spending our money where is it going how are we spending it

28:59 why are we spending it all those things um to make it as transparent as possible.

29:09 And I think that came from learnings from both of us.

29:15 So like for me, I've been in a situation where the company didn't tell us about financials at all.

29:16 And I had no idea how much money, runway, all those things that you worry about in a

29:20 startup.

29:21 And then I was let go because they were running out of money and, and they just didn't have the money to,

29:26 to,

29:26 to pay me.

29:27 Um,

29:28 and it was like a big surprise.

29:30 And I was like,

29:30 Oh God,

29:31 now I've lost my,

29:31 like,

29:31 how do I provide for my family?

29:34 Or how do I provide,

29:35 you know,

29:36 how do I pay for my bills?

29:37 All those things.

29:38 And,

29:38 um,

29:39 so we try and be as open and transparent as possible when we,

29:41 um,

29:42 do that.

29:42 And we,

29:43 we,

29:43 they have unfed access to essentially what we send to the founders every,

29:47 every month. I mean, the investors, not the founders. Yeah.

29:53 I love that. Excellent. Thanks so much for sharing this. And you know,

29:57 the previous points about marketing and inbound and outbound sales,

29:59 like it might be like, you know, kind of like go,

30:02 can go without saying for a lot of people,

30:04 but for people getting started right now, they haven't worked in startups in the past or in developer tools infrastructure.

30:10 And I think this can be very useful.

30:12 So thanks.

30:12 Thanks for elaborating on that.

30:15 And now building a sustainable business, having paying customers, pricing gives people a hard time very often.

30:24 And I mean, i couldn't i

30:27 couldn't stress this more like same same with us uh here at algora so is there something you could

30:32 share about this or maybe a mistake to avoid from what you've seen so yeah yeah yeah so uh so unkey

30:39 when we first did the pricing it was just like i don't know this will do kind of thing like that

30:44 was the attitude we had

30:45 like we just didn't know what's power for pricing we we were trying to understand like we did some

30:50 napkin math about how much each verification cost us per million and like try and work backwards

30:55 from that um and and at first we were like yeah this pricing makes sense um and then what we saw

31:01 was like a common theme of like people would approach us and be like, hey, like want to use your service, but your pricing is a bit skewed.

31:08 This is what we need.

31:10 Is there any way we can work out some sort of pricing for us?

31:15 And we did that one time.

31:17 We did it for our biggest customer because they were like, we're going to do millions.

31:20 And we were like, oh, my God, someone wants to pay us.

31:23 Let's do special pricing for them.

31:24 millions and we were like right oh my god someone wants to pay us like let's do special pricing for them um and so we just changed our pricing as of uh this year uh and essentially what we did was we

31:33 15x the amount of verification to get um we like half the price so basically uh pricing

31:42 is not permanent that's the first thing i want to say is pricing is not permanent. That's the first thing I want to say

31:45 is pricing is not a permanent feature.

31:47 You look at it and you say,

31:49 I think this is what my customers

31:50 are willing to pay us.

31:51 And I think this is fair.

31:53 And odds are you've got it wrong.

31:55 The odds are that whatever you pick

31:57 the first time you put pricing out is wrong.

32:01 And what you need to do at that point

32:02 is then take the feedback from customers.

32:04 Every time someone makes a comment about your pricing, whether it's good or bad, right?

32:08 They're like, oh, it's so cheap.

32:09 Thank you so much.

32:10 Like, you've saved us so much dev time.

32:11 Like, great.

32:11 Okay, that's a pass one for the positive pile.

32:14 But if you continuously get things like, do you offer custom pricing because we need to do this amount of work and it's going to cost us X and that's more than we're willing to pay,

32:27 you take that feedback and you think about it.

32:30 So I spent a lot of time,

32:32 and we actually made this kind of mistake at Clerk too,

32:34 which was I looked at every, we, at Clerk,

32:37 we looked at people in the industry and we basically just copied

32:40 and pasted what they did.

32:42 And similarly, I actually did this when we first did the pricing

32:45 at unki what we did was we looked at authentication providers and said this is what they charge for

32:51 like monthly active users or whatever and we'll base it upon that um the second iteration that

32:57 we just did i actually did it more on like okay if i'm a business and I pay this much amount, how many times can I run a verification against keys

33:08 before I have to pay Unkey more money?

33:12 So like $25 gets you in the door.

33:15 Okay, if I get to here,

33:17 potentially could I be monetized at that point, right?

33:20 So the original pricing,

33:21 it was 10,000 verifications and 250 keys, which sounds like a large number until you actually think about how APIs work.

33:29 And then you realize, oh, no, that's not really a huge number.

33:33 It's actually quite a small number.

33:35 So sooner or later, you're going to have to pay me pretty quickly.

33:38 So what we did was I sat down, I looked at API in general and and looked at like gateway pricing and things like that,

33:48 and movers and shakers in the industry and said, okay, how can we make this as easy as possible

33:54 to say $25 is a bargain. And then you're kind of locked in and you feel comfortable paying us more

34:00 money. But also you feel like by the time you actually have to pay me more money you've already made enough money to cover it um and so i kind of went through a few iterations i

34:10 sent it to the team the team came back and said like nope still think it's too expensive like

34:14 you need to move some stuff around so we moved a bit more and then we landed on pricing is today

34:18 um and yeah and i imagine in six months maybe we'll revisit again and I'll look at it and say, okay, maybe we need to change this or that.

34:29 And, and, or maybe a year from now, but right now what we have for pricing is it feels super fair.

34:35 And I think the biggest thing for us was we changed the pricing and then the next day, uh, the, we changed it on Friday.

34:42 And then on Sunday we had a new customer sign up.

34:45 And then on Monday, they paid us. So for us, that felt like, okay, yeah, that we definitely

34:50 made the right decision here. There was no conversation. I never even interacted with

34:54 this person. They just signed up, signed up their business, and then immediately paid us.

35:00 And that felt kind of like, yep, okay, I think we're close to what the pricing should be.

35:05 And that felt kind of like, yep, okay, I think we're close to what the pricing should be.

35:10 And being self-serve, you don't really want to talk to anybody if you can avoid it.

35:14 Great. Thanks so much for sharing this.

35:16 Again, super challenging topic.

35:22 Being an open source maintainer and, you know, having also Andreas, of course, had his own pains or glory days.

35:27 Is there something we can say about how you're running the process,

35:31 advice to people who might be getting started now?

35:36 Yeah, for sure.

35:38 So I think I'll do two pieces here.

35:40 I'll do contributing to open source as one

35:43 and then being a maintainer as a different piece so if you want to contribute to open source which i advise if you're even interested in just

35:50 like seeing how maybe real products work in the world and it's not corporate based um find people

35:57 that uh are good at labeling things uh pick a project that you're passionate about uh like i

36:03 always cow.com is like the gold standard

36:06 of commercial open source. A lot of

36:08 people know who they are. And if you don't,

36:10 then maybe you should look up who they are and then

36:12 delete your Calendly account and sign up for a Cal.com.

36:14 But anyway,

36:16 they are essentially

36:17 what the gold standard is really

36:20 for open source these days,

36:22 especially in a SaaS-based

36:24 product. There are others out there.

36:26 TRPC is a great example.

36:28 Even Theo's CreateT3 app is a great example

36:30 of other projects that aren't commercial-based.

36:33 Can you solve this, by the way?

36:35 Yeah, us too.

36:36 We're all using fairly similar stuff.

36:39 But the idea here is find a project

36:41 that you're really passionate about

36:42 and go and help them.

36:45 And join their Discord or join their Slack or whatever kind of community they have.

36:49 That's step number one if you really want to start into this space.

36:53 And then step number two is don't be afraid to do non-code contributions.

37:01 Documentation is probably one of the most lacking pieces in a dev tool company

37:06 and even like a spelling mistake sucks in your docs right like if you spell something right it

37:12 sucks whatever doing a pr and and fixing that issue uh is amazing um first we appreciate it

37:21 so much um and i think people think that,

37:25 no, the only way to contribute is code.

37:27 Like, no.

37:28 Please look at our docs and be like,

37:31 this sucks or these words don't make sense

37:33 or whatever it might be, right?

37:34 Or this is out of date.

37:36 Contributing to those really makes a big difference

37:38 to open source companies.

37:40 And then secondly, if there is like a first time issue tag,

37:44 look at those and kind

37:45 of feel out and read the contribution guide so that you know how these companies work.

37:51 Cause some are a little different than others.

37:53 Uh, we have a very slightly different, uh, thing in ours, which I'll talk about in a

37:58 minute.

37:59 Um, but overall, like, yeah, read the contribution guide, start with something simple, like docs

38:03 or something just so you can understand the workflow from end to end, right?

38:07 Like, okay, I've made this change, I've put a PR in,

38:09 you know, I followed all the steps.

38:12 Okay, now that's been, you know, merged.

38:14 Okay, now maybe I feel more comfortable taking a look at the code

38:17 and setting that up and getting going.

38:20 And even up to that point, may I add that, you know,

38:23 you set up the project, you took it for a spin,

38:26 maybe you can open an issue and provide some feedback.

38:28 How that experience is better for the next contributor.

38:33 And similar documentation or even landing pages, example apps.

38:39 There's really an endless list of ways you could contribute code or not code.

38:40 Yeah, for sure.

38:43 And then as a maintainer, right?

38:47 My advice is make sure you put some structure in place and make it make sense for everybody, right?

38:51 So the people at Anki follow the same standards

38:55 as people that would contribute from open source.

38:58 So we have one rule, which is there must be an issue

39:02 associated with whatever the PR is.

39:04 one rule, which is there must be an issue associated with whatever the PR is.

39:12 So if you're a contributor, essentially all it requires is that you open an issue and somebody approves it on the team. For internal purposes, everything is pre-approved. So

39:18 if you're one of our maintainers, it's pre-approved. So you don't have to wait until

39:23 someone says like, yep, that's fine. Go ahead, do whatever.

39:26 For bug fixes, we don't really enforce it,

39:28 but for features and any enhancements we do.

39:31 And then secondly is making sure you have like great issue templates

39:36 and feature request templates and documentation templates.

39:40 So it's easy for the user to get into that space.

39:44 And then similarly with your contribution guide, making sure people understand how your

39:49 contribution workflow works.

39:51 So for us, we have like PRs have to be in a specific format.

39:56 Um, you know, you have to do these following steps before we'll accept the PR.

40:01 You have to sign a CLA.

40:02 All those things are kind of like required.

40:04 And we make sure

40:05 that's clear when you open a pr um and then as a maintainer is keeping good keeping on top of those

40:14 prs that come in it's very hard like for us it's it's it's super difficult like we're trying to

40:19 ship fast and like produce product because we we you know we want more people to pay us and all these kind of things.

40:26 But keeping

40:27 on top of them, so we try

40:30 and do it once a week

40:32 minimum of

40:33 each person in the team can look at them and

40:36 say, okay, yeah, this one's ready. I feel

40:37 comfortable that the code looks good.

40:40 I've tested it. The tests don't

40:42 fail, blah, blah, blah. Everything looks good. We can probably

40:44 merge this.

40:47 And we'll take it for a quick test bin in Vercel,

40:48 and then we'll test the API.

40:51 And then when we're happy, we'll just merge them in.

40:57 And then letting people know that, like, hey,

40:58 this PR has been open for a while.

41:00 Like, you've opened a PR, we've asked for a change,

41:01 and you've not done anything.

41:04 Like, is there something we can help you with?

41:09 Or are you just busy? Do you want us to close this pr like what was the status um on the other end um and then one big thing that i think some people don't like but i particularly like is

41:16 can i be assigned this issue seems to be a controversial topic in open source so like i

41:23 prefer if you ask me if like hey can i take this issue on before

41:26 working on it uh because we found that even if we have people assigned to issues people will

41:31 still try and contribute that issue and it's like well i have internal team members working

41:36 on this issue right now they're looking at it right now um and so having that like just that

41:43 like hey i'm looking at this issue.

41:45 I think I can take care of it today.

41:47 Like, do you mind if I take this issue?

41:49 Really helps us out because we get notified in our Slack that someone's asked a question and we go and look.

41:55 And two, it kind of starts that relationship between you and an open source maintainer.

42:01 Excellent.

42:02 Thanks so much.

42:02 This is super helpful for people to hear

42:05 really appreciate it

42:06 and

42:07 I think

42:08 we can

42:08 move to the

42:10 demo

42:10 of

42:11 if you would like

42:12 absolutely

42:14 my co-founder is up

42:15 joining

42:16 let me see here

42:19 let me present my screen

42:21 and

42:24 maybe just thinking about it.

42:28 There we go.

42:30 Let's talk about Anki.

42:32 First, let me sign out.

42:34 Let's start from the beginning.

42:36 Okay.

42:36 So if you don't know what Anki is, and maybe you do, maybe you don't,

42:41 depending on who's listening to this,

42:42 essentially the idea here is we provide API authentication

42:47 for being able to scale a user-facing API.

42:50 We provide ways to create

42:51 keys, verify keys, manage your keys,

42:54 and essentially you can do it in seconds.

42:56 We're an API-first company,

42:58 so I won't be showing the API today,

42:59 just because I'm on a machine that doesn't

43:02 have any development stuff on it.

43:04 It's just my content creation machine.

43:06 But we can show most of it off in the dashboard anyway.

43:10 So once you sign up for an account,

43:12 you'll get dropped into this kind of screen right here.

43:15 And this is what we call a workspace.

43:18 Now, workspace can either be a personal account

43:20 or it can be for your businesses.

43:21 As you can see, I have many different businesses

43:23 that I've tested over the months.

43:27 And an API is essentially a way for you to group API keys together.

43:31 So this could be production, this could be test,

43:34 this could be staging, or it could be an API route, right?

43:37 It could be like this one is AI fun,

43:41 so that could be a route that you have for your company.

43:45 And then looking inside your APIs, so if we load this API up,

43:51 we give you out of the gate essentially ways to look at how the keys are performing,

43:56 what you've done in the month that you're currently in, and how those verifications work.

44:02 So we have basically three real types of verifications

44:07 that could potentially happen.

44:09 One is what you really want, a successful verification.

44:13 This just means that we verify the key and the key is happy.

44:16 It's valid, it's enabled, all those kinds of things.

44:20 Then we have rate limited and usage exceeded.

44:22 So you could actually see if someone's maybe potentially abusing your API.

44:27 And then the same for active keys.

44:28 We show you how active your keys are over the last seven days,

44:31 and you can push this all the way out to the last three months.

44:35 With this, essentially, the idea here is we only actually charge you

44:39 for successful verifications and anything else that happens.

44:43 So if it's rate limited, it doesn't exist, it's disabled,

44:47 the usage has succeeded, it's expired,

44:50 all those things we actually won't charge you for.

44:53 So that's how the usage-based works.

44:56 And then you can essentially play with your keys in here.

45:01 All of the stuff that you see in the UI, you can do in the API.

45:06 So as you can see here,

45:10 all these keys are enabled. None of them really have any of these features, but I'll show those features off. And then you can see when they were created by who owns it, what the name potentially

45:16 of the key is, etc. So when you create a key in Unkey, essentially you have um for advanced usage and then just the standard stuff so a

45:27 prefix can be anything to make it easier to identify so maybe you just do prod um and we'll

45:33 underscore it and then put the key on top uh then you have an owner which can be anything

45:38 maybe it's from your authentication provider so maybe it's user xXX, you know, like some sort of user ID that you can easily

45:46 identify this owner as. And then the name is just a name of the actual key itself.

45:52 And then you can add all of these custom things. So for example, limited usage,

45:57 you can add that this key can only be used 100 times. And then you have the option to actually refill this automatically. So for example,

46:05 I could go in here and say every day, add 10, and that will refill 10 to this key.

46:12 When you click create, we give you the key here, which I'll just show. This is the key prod

46:18 underscore and then whatever our key is. We only show you this once and there's no way to retrieve this key. So once you've

46:26 issued it, whether that's through your own UI or you're just manually giving this to a user,

46:33 that's the last time you'll ever see the key. So they need to be stored somewhere safe or just

46:39 pass it on to your user and your user can have it as is. And so this essentially is how we verify any key in our application.

46:48 What we do is you curl this endpoint with your key and we'll tell you whether or not

46:54 it is valid or not.

46:56 And I can technically show this using recbin really quick.

47:01 Don't do this in production.

47:04 But essentially what we do here is we'll send the key down.

47:08 And when we run this, maybe not.

47:12 Maybe it won't work.

47:14 There we go.

47:15 You'll get this back here, which is the key itself.

47:19 Is it valid?

47:20 And you can check against that.

47:22 And then the owner ID, any metadata associated with the key

47:25 and how many remaining it has left.

47:27 So right now it has 99.

47:29 So if this user uses it another 99 times,

47:31 the key will then become invalid

47:33 and you'll get a response back

47:35 that says usage exceeded.

47:38 The idea here is

47:39 these are all globally distributed keys.

47:41 So regardless to where the user is,

47:43 you'll get the same minimal latency from start to finish. And then we cache it really heavily. So even when you get the key

47:52 for the first time, it should just be as fast. Now, if we go here and look at created at,

47:57 and this one is our latest one here from Thursday, you can see it has 99 remaining,

48:02 it has a daily refill amount, and the user ID is here. And then you get see it has 99 remaining. It has a daily refill amount and the user ID is

48:05 here. And then you get the options to do a bunch of different things. So if we go to details here,

48:10 we can see when it was last used, success rate, rate limited usage exceeded, when the verification

48:16 was, where the user agent was, what the IP address was, what the region was, all those things

48:21 are provided. And then you can actually update these keys. So

48:25 through the API, which I'll show in the docs in a second, you can update these keys and also through

48:30 the UI. So if you decide that remaining is no longer required, you can just disable it, hit save.

48:36 And now this is no longer on and this key can be used however many times you want.

48:39 And vice versa, you can enable it and then set that to whatever you may need.

48:43 And vice versa, you can enable it and then set that to whatever you may need.

48:57 The idea here is that we want you to be able to essentially handle any scenario at any point without having to reissue a key or make some changes in your database to make this work.

49:07 On top of that, we have this new audit logging system which is just new it tells you who's creating keys when they were created all those kinds of things so that you can look at these and say okay

49:13 this API has been created who created it and maybe you can go and ask them why

49:19 and that's obviously internal you can ignore the success thing here. That's just for me.

49:26 And then that's essentially it.

49:29 And then we have a bunch of stuff in here

49:30 for things like if you want teams

49:32 on a non-personal account,

49:35 you can invite users.

49:37 Root keys are essentially

49:38 what drives all the resources.

49:40 So when you create a root key

49:41 for the first time,

49:43 you can now actually create keys via our API.

49:47 And then you can use that to update keys, delete keys, revoke keys, all those essential pieces that you might need.

49:55 And then I can just quickly talk about the docs because docs are probably going to be pretty important here.

50:06 important here. So if we go to our API reference here, you can see that we have

50:11 the ways you can actually test this in the documentation if you so wish, but it shows you in all different languages how to expect this using just the straight

50:18 HTTP requests and what is required and what is not. And then we have libraries in some of your favorite languages,

50:27 whether it's TypeScript.

50:29 We have a TypeScript one, which is an official SDK that we built.

50:33 Same with the Hono package here.

50:35 So if you're a Hono fan, you can use this to do middleware verifications.

50:39 And then our contributors, such as Python, ElixirGo, Noxt,us REST, Spring Boot are all things that you can use

50:48 and not have to worry about having to turn HTTP requests

50:52 into the language that you choose. Of course, we have an open API

50:56 spec, so that makes it fairly easy. If there's one that's not available,

51:00 you can use that to drive this information.

51:04 And then on top of that if you're

51:05 interested in contributing we have a guide here that shows you how to set up everything in our

51:11 application it tells you what's required and then the rest of the third-party services

51:16 are just optional we show you how to do versioning we show you how to add some features if you want what we do every day.

51:26 And that's essentially Unki in a nutshell.

51:31 We have a bunch of different features that I probably should just go through real quick.

51:35 So we offer you rate limiting.

51:36 We offer it in two ways.

51:38 First way is like local.

51:40 So it's super fast and on the edge.

51:42 And then we have this global consistency version.

51:44 So if you're

51:45 really conscious about having a strict rate limiting and you're not too concerned um if

51:51 you're really concerned about how it and it should never be exceeded you can use this uh global

51:56 version which adds uh latency to our verifications because we have to go and check our database every single time. But the fast on the local side, which is edge everywhere,

52:09 it will be much, much faster.

52:12 Temporary keys, which is essentially a way for you to just say,

52:14 hey, it expires at this time.

52:16 So if you need a monthly or weekly key or some variant of that,

52:20 we can offer that out as a way to basically expire the key.

52:23 Remaining is what I showed you already,

52:25 which is essentially just like, hey, 100 times,

52:28 and that will be how many times you can use it.

52:31 Great for things like AI.

52:32 If you need to purchase like 100 uses of some sort of AI API,

52:37 you can use that.

52:38 Obviously, refill is part of this remaining feature.