#Kiro Pricing Feedback

1 messages · Page 2 of 1

winged harbor
#

My second feedback about pricing.
Kiro sent me a message that the system will be free until mid of this month.
I don't need free service. I am currently paying a subscription elsewhere and happily added credit after I used all allowed credits in another platform, which lasted me an amazing 2 weeks of work, with good quality results.
Service should be only free and very limited only for a day or 2 of testing for the user to make sure the system does the job it claims to do. Then it's normal to have user pay. But you really have to understand I am looking for something viable long term, with affordable pricing for usage long term. Not free 15 days then crazy expensive 15 days every month or I don't know what is the plan for after the 15th of the month. When I get a month subscription, I expect the credits to AT LEAST, last 10~15 days, depending on the subscription pricing. And not 2~3 hours.

jaunty grotto
#

Kiro is a waste of time. You will complain about pricing for the rest of your life. The fact is that the usage we want is only somewhat provided if you go all in on 200 a month. And if you are doing that, and not offering a real subscription ( i.e. a month means a month usage ), the only workable compromise is one similar to one already offered by claude code directly, which is a rolling 5 hour session quota, a weekly limit, and monthly limit at around 200 a month. The difference is Kiro is only offering Claude Sonnet, and Anthropic also of course offers claude 4.1 Opus. Unless Kiro can come up with a MORE generous usage quota than Claude subscription does already, there's simply no point to this project.

muted lake
#

A pricing hack for non spec users. Keep updating task list manually and add different texts there. This will consume spec requests, not vibe (at least the first invoke). You can use your spec requests this way until prices gets fixed.

Basically instead of texting on the vibe section you just text as a task 😆

ashen holly
#

For "Session Too Long", press the button "Sumarize", Kiro opening the new converation, with "Understood. I have the context from our previous conversation and am ready to continue helping you." and 1 interaction from my part = 7 vibe credits (1 VC)

jagged topaz
#

don't feel bad I just used 23 specs on 2 tasks

#

3 vibes for summary

#

I have an extension that put a dollar sign on that button to prevent me from using it.

#

Grok4 coder is a ninja at fixing kiro typecheck errors.

jaunty grotto
#

I have given up on Kiro at this stage. I think its about 1 year away from becoming usable, and only then if the pricing is competitive and 1200+ bugs are fixed, and billing is available everywhere, and missing features like BYOK are enabled. I was hopeful initially, but there's just too many issues there, and unless they listen and fix them, I'm out. Good luck all.

candid copper
jagged topaz
candid copper
#

because its free

jagged topaz
#

we will see how it holds after the 10th

jaunty grotto
#

The merit of Grok Code Fast 1 reminds me instantly of the XGH ( Extreme Go Horse ) approach to coding:
In a nutshell:

  1. I think, therefore it’s not XGH.
    In XGH you don’t think, you do the first thing that comes to your mind. There’s not a second option as the first one is faster.

  2. There are 3 ways of solving a problem.
    The right way, the wrong way and the XGH way; which is exactly like the wrong one but faster. XGH is faster than any development process you know (see Axiom 14).

  3. You’ll always need to do more and more XGH.
    For every bug solved using XGH, seven more are created. As all of them will be solved using XGH, then it's utility tends to the infinite.

  4. XGH is completely reactive.
    Errors only come to exist when someone notices them.. ( Full explanation here: https://medium.com/@dekaah/22-axioms-of-the-extreme-go-horse-methodology-xgh-9fa739ab55b4 ).

Medium
22 Axioms of the eXtreme Go Horse Methodology (XGH)

I think, therefore it’s not eXtreme Go Horse.

mortal wharf
#

I'm testing the $3 Z.ai API with the Claude Code app. Pretty decent. Super fast.

Gave it my Kiro Specs and told it to work on two tasks. The stats when I closed Claude said cost: $173. Not sure if this would be the cost in Claude API, but I'm glad I'm not on it 😅 .

For the first two tasks on this Spec, Kiro took 12 spec requests from me => $2.4 at the overages pricing, almost what I paid for the whole month on Z.ai (this cannot be sustainable, right?).

I'll try to make it work on all other 20 tasks for this spec, in the next days and see how it goes...

dusky yacht
spice laurel
tame flint
#

Quota allocation is not sustainable. Consuming 100 vibe and spec requests in less than 24 hours, mostly due to retrying tasks that faced system issues, makes me question what tier I need to be on to use Kiro effectively for an entire month. The overages seem like a fast track to debt, given the current usage for requests.

winter sluice
#

Ran out of free usage and trying to upgrade with 20 dollars monthly payment. My country is not in the billing address list. Which country should I choose ? If I choose randomly from the list and pay the price, will I get refund ? Can somebody help me?

hybrid furnace
#

They should just use a 1 credit system instead of 2... this allows users more flexibiltiy if they do not require specs or only needs a fraction of it

loud pagoda
#

i paid for the pro plan on kiro today and i got fewer spec and vibe request. If you look at the second image from kiro website you will see that it says i will get 225 vibe request and 125 spec request but i only got 190 vibe request and 106 spec requests in the first image. This is not what i paid for. What is the solution to this kiro team?

Screenshot_2025-09-05_185824.png Screenshot_2025-09-05_185747.png
hybrid furnace
loud pagoda
#

no i paid 16$ plus tax which equals to 20.22$

loud pagoda
quiet dew
#

Why is the pricing so complicated to understand? Everybody knows and understand the concept of tokens or messages or requests. Why try to innovate with pricing structure? I understand nothing. Moving to Qoder.

loud pagoda
#

is qoder better than Kiro? I tried qoder but every chat session have a shorter context than Kiro? What do you think?

#

and it still have vscode old UI

mortal wharf
#

Kiro basically 🦶 🔫 itself with this pro-rated thingy. The PM decided to charge credit cards and reset the credits of all users in the first day of the month, instead of keeping an extra field in the database to store when the plan should be reset for each user. This is causing FUD from the very first day people started to see less credits in the account than what we see on the plan page. As the only way to see the 125 credits is to sign up for the plan on the September 1st, or wait for the next month to arrive.

hybrid furnace
mortal wharf
hybrid furnace
muted lake
#

I guess new deployment happened because right now every task consumes both spec and vibe requests. It used to be that it first uses a spec requests and if you chat it switches to vibe. 🤷‍♂️

hybrid furnace
muted lake
hybrid furnace
vernal sable
#

I see people are fed up writing here 😁 Very less messages here now days

half hill
#

I don’t fully understand how vibe vs spec is counted. I consumed a spec token when I created a spec to do some work on my app, that’s fine, but as soon as I start one of the created tasks I consume another spec token in addition to vibe token. Then when the thread becomes too long / out memory I then have to continue in a new chat. When I continue in that new chat I consume another spec token?

#

Here is my current usage and I have created 14 specs during this period.

IMG_4719.jpg
vernal sable
#

If you have hooks, turn them off. Because hooks makes it worst. After that the frequency of vibe request consumption will be far more less. But it will be there. 1 request might consume additional vibe request depending on the complexity as I have observed. It is incorrect too. It must count only 1. That's the purpose of counting requests. Otherwise we will never be sure how many vibes will be used for your single request.
Spec is also counted similar way.

#

I really hope that KIRO team has listened to these feedbacks and solve the issue in next pricing release.

strange seal
#

try warp.dev, great value and priced correct and it is fast with mode model options. I switched from Kiro to Warp and I cannot be happier. I will use what is left with my sub, Kiro was a diapointment with the pricing and bugs anmd performance. it is a shame, they had a winner and they blew it. With Warp now, you can try 2500 rquest for $1 promo, just search for Warp on you tube and get the promo code and try it.

harsh iris
#

the current code is MOO

#

until they change it

wicked kestrel
#

Can anyone provide feedbacks on warp?

jovial jackal
#

Why don't I see the overage refund? I've got an email that the overage up to 1000 vibes will be reset

Screenshot_2025-09-15_at_8.57.21.png Screenshot_2025-09-15_at_8.52.37.png
signal imp
harsh iris
#

i've had Sonnet 4 eat up f.e. 20 requests for one prompt which did more stuff

#

putting Warp aside, today's Kiro pricing announcement is worth waiting for 🙂

river rose
#

When is "today"? I'm hesitant to continue using on project work until I know what I'm actually being charged

harsh iris
strange seal
# signal imp How does warl count it's request? Is it similar to Kiro or windsurf?

yes. it is stright forward. a request you submit to the AI and it will process. the request/price is just great, Add to it the stability and consistancy of Warp . I used Kiro for a while (since it came out) and I could not keep up with the IDE bugs and issues, I hope they fix them. I did pay for the 200$ plan for kiro (I code 10 hours a day) and I caneled it, I just couldnt deal with the bus and issues the latest versions produced. I have created a ticket in Kiro-help a week ago and yet for anyone form Kiro to answer me or give me any suggsstions or solutions.

river rose
#

Oh yeah I forget time zones 😅

harsh iris
#

Also, announcement won't necessarily mean they implement changes starting today

lusty igloo
harsh iris
#

You know I can't confirm or deny 😉

vernal sable
#

Pricing page remains the same. I don't have any hope that they have improved on pricing. Let's see what happens tomorrow.

midnight spindle
placid remnant
#

Free until October 1? ☺️Does that mean that I can continue using overages without being charged with my pro subscription or can/should I upgrade to higher tiers? 🤑

hybrid furnace
#

#announcements message

Are we gonna get refunds or payment waiver?

mighty bane
#

Yes. It's all in the FAQ

pulsar crest
#

I gues its ... Lets see on Pricing - IMHO at this point i am just going to pay for pro and see if there is worth it over just using VS code and Claude Code CLI

maiden field
#

Not sure what happens to my existing used vibe/and spec usage count .

Screenshot_2025-09-15_at_12.13.15_PM.png
timid sigil
native acorn
#

Noticed the new pricing plan. $20 for 1000 requests. That is OK IMHO. But, please don't rush to introduce another "Auto" thing. I mean, for the pricing plan / usage itself, we have been ding-dong-ding-dong for quite a few rounds. You guys need to cool down and wait for more feedbacks before introducing a new "variable" in this already-complicated usage scenarios.

half hill
hybrid furnace
#

So are we gonna be charged for our current overage or how's that gonna be like

image.png
plush sonnet
hybrid furnace
vernal sable
#

Any one tested this new pricing? How long it lasts?

muted lake
#

I see simple prompts now consuming multipe vibe requests, given that they made it 1k requests for 20$, im still not sure if they really improved anything or simply kept the broken logic but increased the limits. Couple of hours of play today and already 50 vibe request is done and couldn't even get a simple feature done.

So i suspect 1k still won't be that good but a bit improvement i guess

slow nebula
#

When will the new pricing be active? On kiro.dev it says "coming soon".

fresh halo
#

Was hoping to test "fractional" common credits before deciding for paid plan, but still has Vibe/Spec requests on my free account

timid sigil
muted lake
# timid sigil Out of curiosity, how much work were you able to accomplish in those two hours? ...

I think quality has decreased significantly. A simple feature costing 100 vibe requests and barely getting it done. I actually started doing more of copying code and using chatbots somewhere else.

I think output quality was bad compared to the cost. Many times I tell agent to not do any code changes and I wanna discuss architecture design before but it starts implementing stuff already.

40 hours is really nothing if nothing is produced at the end. I started doing more manual work then telling kiro to do something simple.

I guess pricing - quality is really tied together here. It'd be okay if the work was done in 30 mins (10 mins work for software engineer) but not in 2 hours

primal swallow
#

+1 on what lorsofallen said. I'm resorting to using other tools like Codex CLI to do the rest or fix bugs it cannot (or doing it myself, but then I'm not vibing...).

jade tartan
#

Pricing could be same, but unified as one usage along with a few more models to be integrated.

vernal sable
#

Why is session summarization takes 3 vibe credits?
I think Kiro just did a blunder again. They just eplaced the term request to credit and served the same recipe to us with the new name. Not sure what's the differrence Auto mode makes.
5 spec tasks consumed 10 specs credits.
Generating Requirement, design, tasks documents consumed 7 credits.
Session too long summarization took 3 vibe credits.
Rest 19-7-3=9 credits not sure where it went. No account of it.

#

It is understood if what we use is accounted. But when there is no account where the credits are going, it's difficult for KIRO to be widely adoptation.

#

I heard that there is Memory Leak concept in software development term. Here, it is credit leak. 🤣

lusty igloo
# vernal sable

Same here still stuck on this separate counting for vibes and specs

primal swallow
lusty igloo
#

aw man
I've been waiting for it to upgrade so I can use further

vernal sable
#

Kiro IDE still follows the same pricing model. They’ve simply swapped the word credits with spec and vibe requests. In reality, nothing has changed—it’s just the same recipe with a new label.

It’s like going to a restaurant, sending back a dish you didn’t like, and expecting something freshly prepared. Instead of cooking a new meal, the kitchen just swaps the plate and serves you the very same food you returned. You feel satisfied because it looks replaced, but really, it’s only your ego that’s been fed.

#

Still spec and vibe requests not credits. This proves my point.

image.png
#

This page says something else. Eventually confusing the developer community.
Why can't KIRO treat every user in the same way?
If at all they have changed anything backend side, why treating existing users and fresh users differently?

image.png
harsh iris
#

I'd assume some announcement incoming before the new billing cycle hits in

steep stream
#

Needs way more perks

image0.jpg
vernal sable
#

Today, new credit system was visible in my IDE. Let's see how it goes. I will test it tomorrow to check if is better than spec/vibe.

mystic drift
#

Received a about a week ago, liking it better so far. But spent mostly bonus credits.

#

I was very frustrated with having remaining specs while paying overages for vibe

proven orbit
#

Currently pay the $200/m plan for Cursor and comparing both it seems like I consume my credits a lot faster making it seem like I have less usage included in my plan with Kiro and more with Cursor. The bonus with Kiro is that with the spec driven dev it seems to accomplish a lot more than cursor does with those spent credits while making less mistakes. I definitely like the feel and approach Kiro takes over Cursor. The big take away will be if I can get more or the same amount of work done with Kiro for the same price as Cursor before having to pay overages.

dusky yacht
mystic drift
#

Yep, received about a week ago. That was my point. I was frustrated, now i'm not.

obsidian lagoon
#

I'd say it does feel like credits are spent quickly, but the unified credits is nice for tracking/forecasting.

One of the things I noticed is that I used sonnet 4.5 at work extensively with a dedicated key and blank check. Tracking my usage in cline and comparing against similar work in Kiro, it doesn't feel like being part of the kiro ecosystem is actually saving me that much if any on AI usage costs in aggregate. Cline shows API call pricing for each request and for each thread so it was easy to see the net.

I can't prove it but it feels like 500 credits amounts to about 10$ of sonnet 4.5 which would represent a linear pricing model that also is in line with the linear packaging of 1000, 2000, 10000 credit monthly allotments.

I would think that kiro would be able to pass on those savings in an obvious way being part of AWS (I know business doesn't always work that way and the team is likely subject to the same access rules as everyone else) so we see a discount as packages move up in price.

I'm tempted to move to the power package if I knew that I'd be gaining either 1 to 1 access to frontier models without the premium (1.3x) or if I knew I got 13,000 credits (1.3x) anyways so there was again a 1 --> 1 relationship on my premium "power" membership.

Open to thoughts on this, but I think this represents a clear pricing to experience alignment

proven orbit
obsidian lagoon
# proven orbit Yeah, I’m on the same page. The unified credits are nice, but the actual value j...

My other reason was that I just disable all automatic hooks and develop alongside kiro to prevent any waste or runaway tasks, so unless I felt comfortable with my credit allotment being a good investment in my overall development costs (power = more scale) I'm not going to scale up with automation and the power tier. Because 40-->200 is enough that we're talking about ai development costs being disproportional to saas infra/hosting costs, that kind of thing.

steep stream
#

all plans run exaclty the same you just pay for more credits

#

its crap

#

unless top tier gets more perks im out

steady stag
#

I see kiro folks are still discussing the same stuffs that I've been shared since the first pricing presentation. Since then, I see improvements however it seems it's still not enough. I see like 50% less token consumption and can be enough to make people happy if it gets 50% lesser. Then the pricing and token consumption will be competitive enough. On the other hand, highest packages are still not charming.

The main point is there are many competitors that they are investing for future. Like windsurf, qoder, trae and many of them. So, I don't think it's a good time to make the highest profit from the product. It's time to improve the developer habits on kiro.

Those were my honest opinions. Let's see!

jaunty grotto
# steady stag I see kiro folks are still discussing the same stuffs that I've been shared sinc...

Yeah - I was probably one of the very first to use Kiro, literally the moment it was available I had added myself to waitlist and got access very quickly. I was initially blown away with the EARS spec approach. Because it is clearly a good idea. But there are also many reasons I cannot choose Kiro if Im having to pay for it. Its not the best out there, it has some nice touches, but ultimately falls short due to some core issues:

  1. I cant even pay for it, I'm in LT, they do not offer payment for LT and many other EU countries in fact. Very strange.
  2. WSL compatibility -> its useless. My project is running on WSL2, kiro cannot use it properly
  3. Lock in to Anthropic models only. I get it, they have a deal with Anthropic. But as good as Claude 4.5 Sonnet is, in my experience, GPT-5-Codex is still better. If I cant use it in Kiro, why should I pay a max plan for Kiro? But still, kudos to the Kiro team, they did create something original, and it is something that could evolve to a leading IDE in the future.. but only if they address these shortfalls.
fresh halo
#

My reevaluation of new credit system. It is much better now, I like fractional credits. Using "Vibe" mode, which is analogous to other IDE, I think it provides better value than Cursor at $20 (price wise), but is more expensive per credit than Windsurf (subjective feel). All changes with "Spec" mode - while it now consumes much less credits (only 2-3 per task, previously 4-5 times more with Spec requests), the enormous number of tasks per spec make it about 10 times more expensive than "Vibe" mode. As example, adding new checkbox setting to GUI and backend (new websocket message, etc) was splitted to 20 task, that mostly modified same 2 files - costed 50 credit without Test tasks. With "Vibe" I suspect it would be 4 credit, for comparison in Windsurf 2-3 credits.

harsh iris
#

@jaunty grotto

Yeah - I was probably one of the very first to use Kiro, literally the moment it was available I had added myself to waitlist and got access very quickly.
That's... not the experience I remember.
For first 2-3 days after Kiro debut there was no waitlist yet, so everyone could just create account and keep using it even after waitlist got implemented.
And those who were unfortunate enough and had to sign up for the waitlist, waited at least 2-3 weeks (if they signed up as one of the first), cause the access code started being sent out atthe beginning of August

unborn patrol
#

when do the credits reset?

harsh iris
#

this is confirmed by Kiro staff

civic trellis
#

can i ask a plan for $10/500 credits for amateurs?

harsh iris
#

Yes you can! ❤️

#

There's been some users asking for the same thing in the past, so hopefully the team will consider it

onyx dove
#

Ig kiro's dead by now, anti gravity is gearing up for a full release and knowing google's practices especially wrt competitive pricing it will be the final nail in the coffin. I have used the product in the early stages, found this one to be different but it all changed when pricing came into effect. The team should've chosen to give generous credit as opposed to making profit from day 1. Since they were the first to have spec driven development (a mere gimmick but works on large codebases) doing so would've amassed users. They could've made more profit from enterprise customers offloading some of it to indie dev packages.

Now that everybody has spec driven development I feel team kiro lost their advantage.

This will all boil down to pricing at the end of the day. P99 vibe coders are Indies and I don't feel kiro represents them anymore.

jade tartan
#

Honestly, if antigravity's paid plan comes up I might switch coz of the support.

tame barn
#

Let’s be honest - 1000 requests sounds generous until you realize each prompt consumes 7–10. You need multiple iterations just to get basic functionality working, plus extra prompts to debug the AI’s edits. It disappears fast, and so your money!

regal echo
#

I'd like to request a plan between Pro+ and Power: $100 for 5000 credits. There is a huge jump between 2,000 and 10,000 credits in those two plans.

harsh iris
#

same users who say today that Kiro is dead by now, will be saying in a few months that Antigravity is dead by now

fluid roost
#

I still dont understand why the plans now are simple multiply:
Pro 20$ - 1000 credits
Pro+ x2 - 40$ and 2000 credits
Power its Pro+ x5 - 200$ and 10000 credits

I even dont mention that the amount of creds are low based on the consumption.. why you just simple multiply it? do some marketing.. let us want to but the higher plan..
do something like this:
Pro 20$ - 1000 credits
Pro+ 40$ and 3000 credits
Power 200$ and 20000 credits

then lower the prices per model.... first do some plan improvements

harsh iris
#

i kinda don't like the system where higher plan gets cheaper credits tbh

#

because it even further makes vibe coding a pay to win game

#

with human coding, the more you learn, the better you are

#

with AI coding, it's the more you pay

small plank
#

i'm honestly looking at having multiple Pro+ accounts with overages turned off

fervent owl
sour nova
#

Whatever the pricing point kiro sets, auto mode should allow free slow usage even if the credit limit is reached. Just like trae and other ide.