I'm looking to migrate off postman now and I have downloaded Milkman and ... oh boi. The only app that has a license to ship a UI like that is Calibre.
The UI looks a bit outdated, it gives me vibes of some old software from 2000s somehow (probably because of the Java).
And it would be really helpful to clearly state benefits of using Milkman instead of Postman right on the home page (e.g. performance benchmarks, UX, extendability, etc.). I see that it has some users and 1.3k stars so there're definitely some cool things in Milkman, probably you could just present them better
At this point, "outdated" is a compliment, it means that it has a menu with standard shortcuts, discoverability, accessibility. It means that it won't take 800MB of RAM to provide minimal functionality. It will stay the same across the years and not be subject of experimentation by UI designers thinking that they know better. Postman's UI started so-so and just went worse from there. SoapUI before it offered similar functionality using Swing. It's always the same app, you can rewrite it but there's no reinventing it.
Maybe this is what and how it always should have been. My life is better now that an LLM proxies my human sound waves and finger clacks to bits that do what I say - and precisely what I say nothing else
It's wild that building bespoke software that fits your needs instead of relying on a one-size-fits-all approach that you're constantly battling against is frowned upon.
I have to ask myself what business you could possibly be in where you look at the software landscape and think "Yea, this aligns perfectly with my needs".
Right, but Postman would have been a valid choice all the way up until it wasn't. The cost of building your own used to vastly outweigh the cost of having to possibly move to another app, pay for licensing when that changes, etc. But that's changing (especially for something like a GUI to make http requests which feels pretty mechanical for basic use cases).
Tangentially related, one belief I've had since the beginning is that LLMs will result in the decrease in large consulting/software factory companies (good riddance) and one-size-fits-all solutions in favor of solutions tailored to a business's actual needs. I work in a slice of healthcare. There are a handful of big EMRs and they were all built 20 years ago in an industry that changes and so many providers end up having to retrofit their businesses around whatever quirks/workflows the software has. And it's usually one of a few flavors: overly restrictive, too configurable, or worst of all: not restrictive at all. None of those are ideal. But now that the economics are changing, hiring a software dev or 2 at your small/medium-sized company becomes a lot more appalling. And it means they actually get business experience. Which means better software all around. I believe the days of shuffling changing requirements through 4 layers of PM's and BAs and sprint leads only to end up with a solution that looks like it was built playing the telephone game are waning.
Are API testing UIs really relevant anymore? I haven't manually tested an API for almost two years, and my APIs are SO much better. This kind of GUI is going to die over the next 5 years.
No, not really. It's already unit and end-to-end tested.
But sometimes the tester needs to make a test where they create something via the official API to use it in other parts of the app. They just open Postman (soon to be replaced), select the collection, select the server, select POST-item and of the the races. Not everything needs to be AI.
We could vibe code something specific for our needs, but that would be more or less again just end up something like postman again? At least I currently don't have the vision to see a big change.
I did, and you really should take a step back and ask an LLM how it can automate what you're doing manually as described above. You'd be surprised what's possible now.
You're not wrong - I bought a Bruno "Golden Edition" for $9 two years ago pre-release... and... I just don't use it. I'm not even sure why I keep it around, at 524.5 MB
That's an interesting point. I wonder if over the next X years consumers will free up ALOT of disk space by just using agents and not needing to download and install desktop software.
Full disclosure, I'm building a desktop app called DB Pro [1] which is a modern desktop database client.
We're still doing well, and selling a lot, but we're having to think about what the next 2/3/4/5 years look like...
I'm thinking the same thing, I haven't have a need for a tool like Postman in a long time, prompting is enough at this point. I wonder what features others use from Postman that an agent can't do quickly.
Quickly eye-balling onprem responses, skipping over UIs and just eyeballing data in general is really useful. Curl can do it but with a couple headers it becomes messy, also RPCs, websockets, etc. It’s just convenient to have a single place where you can eyeball all comms.
This in fact makes me wonder if there is a good “operations center” not just “dashboard” but for all the dev things, websockets triggering, db stats, running queries etc. We have many semi-custom onprems, they do not behave properly as a fleet I cannot just act like they are cattle. These servers contain many products in many configurations. It’s all quite messy..
I find the CPU/disk space dashboards useless. Stuff like monit is useful but status of literal OS services is only part of the picture. There are many layers here and I sometimes wonder how others do this. Surely not all companies just have disposable fleets of “serverless” things?
I don't understand why REST clients are so overcomplicated... We only need an interface for cURL, we don't need cloud synchronization, creating servers, grouping by namespace (or environment, or session, or call what you want), electron, JVM and so on. A simple native app would do the job.
I always feel like some sort of crude brute when I read comments like this. I appreciate a good, modern UI/UX, but will always choose a tool that works well and is FOSS, despite any "last decade" look.
Several years ago when I was developing HTTP/RPC apps, I found MilkMan and never used Postman again.
These days I mostly ask my coding agent to test the APIs. It can use curl create the request, attach all the cookies, whatever is necessary, and send the request, look at the response to all the testing I need. And if I need to manually test something, I ask it to give me a curl command that I can run.
I switched to Yaak, initially compiled it myself but now paying for it. I've not had any issues with it, does what I need and keeps out of the way. Simple yaml files in Git is also nice. I would recommend giving it a spin. I think there is a trial, and you can build it yourself, with a bit of configuring.
I would stay off Insomnia. A few versions ago, they practically took my local configuration ransom so they could force me to create a cloud account. Bruno is (still) not there, but the maintainers are slowly heading that way.
Insomnia has been around for a while but just like this post from a new account, there is a lot of push for bruno, which is usually a redflag. Never trust things that are "preached" for free.
These kinds of things are prime candidate for vibe coding; I don't understand why I would use someone else's. I have a native swift postman ripoff which hardcodes exactly the APIs I'm interested in, and is full of convenience features that make sense to nobody but me.
Sad for all the better postman alternatives, but very happy that postman is in trouble.
How about writing it in C++ or Rust and doing away completely with huge memory footprint and long startup times. Java also needs a runtime, which is bothersome.
I suggest using wxWidgets or Qt so you get multi-platform support for free.
When I saw the UI, my first thought was "is that written in Java"? And indeed, it is. Kinda funny that it's so easy to spot that language based on its horrible default UI skins alone.
The author appears to be suggesting that a key difference between Bruno and Insomnia is that Milkman not written in JavaScript/Electron, but rather JavaFx (i.e., Java) and thus is faster.
No excuse for this in 2026 tbh - get Claude to implement a dark and light mode in tailwind and pick a colour theme. It doesn't mean the product is trash but it's a poor first impression.
1. Very similar UX to postman (just minus the bullshit), so a very easy transition.
2. It's FOSS*
3. Most importantly of all, it has a cute dog logo.
* oof, just saw this on another comment: https://royathan.com/blog/the-api-client-hunt/
Oh boi
I'm looking to migrate off postman now and I have downloaded Milkman and ... oh boi. The only app that has a license to ship a UI like that is Calibre.
I have to ask myself what business you could possibly be in where you look at the software landscape and think "Yea, this aligns perfectly with my needs".
Tangentially related, one belief I've had since the beginning is that LLMs will result in the decrease in large consulting/software factory companies (good riddance) and one-size-fits-all solutions in favor of solutions tailored to a business's actual needs. I work in a slice of healthcare. There are a handful of big EMRs and they were all built 20 years ago in an industry that changes and so many providers end up having to retrofit their businesses around whatever quirks/workflows the software has. And it's usually one of a few flavors: overly restrictive, too configurable, or worst of all: not restrictive at all. None of those are ideal. But now that the economics are changing, hiring a software dev or 2 at your small/medium-sized company becomes a lot more appalling. And it means they actually get business experience. Which means better software all around. I believe the days of shuffling changing requirements through 4 layers of PM's and BAs and sprint leads only to end up with a solution that looks like it was built playing the telephone game are waning.
Postman should be seriously worried.
We have a postman collection mainly for testers to manually verify certain aspect of a otherwise fairly unit and end-to-end test covered API.
But sometimes the tester needs to make a test where they create something via the official API to use it in other parts of the app. They just open Postman (soon to be replaced), select the collection, select the server, select POST-item and of the the races. Not everything needs to be AI.
We could vibe code something specific for our needs, but that would be more or less again just end up something like postman again? At least I currently don't have the vision to see a big change.
This.
Why do you have a collection that is passed to humans for manual testing in the age of agents?
...and... it's gone
Full disclosure, I'm building a desktop app called DB Pro [1] which is a modern desktop database client.
We're still doing well, and selling a lot, but we're having to think about what the next 2/3/4/5 years look like...
[1] https://dbpro.app
Trading bits you own for rented redundant compute to recreate same seems... suboptimal?
How much power should we burn on everyone's agent recreating curl?*
* yes, agents have tools, but what is and isn't an agent tool is a gray area. "Desktop software" could be an agent tool with a human UI.
I find the CPU/disk space dashboards useless. Stuff like monit is useful but status of literal OS services is only part of the picture. There are many layers here and I sometimes wonder how others do this. Surely not all companies just have disposable fleets of “serverless” things?
https://github.com/postmanlabs/postman-app-support/issues/14... https://news.ycombinator.com/item?id=50354960
Several years ago when I was developing HTTP/RPC apps, I found MilkMan and never used Postman again.
[0]: https://marketplace.visualstudio.com/items?itemName=humao.re... - rest client storing config locally
[1]: https://marketplace.visualstudio.com/items?itemName=Christia...
[2]: https://marketplace.visualstudio.com/items?itemName=prawel.r... - same as [0], but with GUI
[3]: https://marketplace.visualstudio.com/items?itemName=slverma....
etc.
Sad for all the better postman alternatives, but very happy that postman is in trouble.
I suggest using wxWidgets or Qt so you get multi-platform support for free.
Went with Insomnia instead (which IIRC also wants an account now?) and now just some text-mode VSCode http client.
Why all software needs to constantly evolve and grow with features? This is how you end up with what Postman has become.
Jokes aside, why is it better?
https://news.ycombinator.com/item?id=50354960
No excuse for this in 2026 tbh - get Claude to implement a dark and light mode in tailwind and pick a colour theme. It doesn't mean the product is trash but it's a poor first impression.