Kagi Assistant standalone subscription

Vlad

(without reading the entire thread) The way I see it a PM at Kagi surfaced an idea we are considering, to get user feedback, something we did countless times before. We haven't decided yet what to do with this, hence request for user feedback.

One thing to understand though is that Kagi has different unit economics than VC funded companies. We can not afford to sell a product at loss knowing there will be fresh VC money in the next round, so who cares. We have been and remain 100% user funded. So any discussion and considerations we have on this front is a part of an effort to keep Kagi alive, not as a part of any 'enshitification' scheme.

ph

Vlad

(without reading the entire thread)

Here's a summary (by Kagi Summarizer, of course) 😉:

Kagi is considering changes to how Kagi Assistant is available on its Pro and Starter plans as it prepares to launch a standalone subscription for the Assistant.
Currently, Pro and Starter plans allow usage of Kagi Assistant up to an AI cost value included in the subscription. The proposed change would limit this to a trial with a fixed usage cap, requiring a separate subscription for continued use.
The goal is to create a cleaner product structure where search and assistant are independently usable and payable, with the Ultimate plan acting as a bundle.
Existing subscribers are concerned about this change potentially being a downgrade and have raised questions about usage caps, plan clarity, and pricing.
Luis from Kagi acknowledged that the transition needs careful handling to avoid negatively impacting current subscribers and that the exact path forward is still being worked out.
Users are seeking clarification on whether features like Kagi Quick and Kagi Research will be part of the new subscription and how usage caps will function, especially for annual plans.
Some users expressed concerns about Kagi being perceived as an AI-first company and the potential impact on its user base.
There's a discussion on whether a standalone Assistant subscription would attract enough users without Kagi Search, with some users noting they don't use the Assistant for tasks like coding or image generation.
Comparisons are being made to services like OpenRouter, with users suggesting Kagi should decide if it wants to cater to advanced users with pay-per-token options or average users with message-based limits.
Users are requesting clear communication about the changes, with some suggesting a 'legacy plan' approach for existing subscribers or the option to purchase additional tokens instead of a fixed cap.

phunkytech

Vlad Chiming in to say that the conflicting messaging and vague descriptions are muddying the waters.

Vlad One thing to understand though is that Kagi has different unit economics than VC funded companies. We can not afford to sell a product at loss knowing there will be fresh VC money in the next round, so who cares. We have been and remain 100% user funded. So any discussion and considerations we have on this front is a part of an effort to keep Kagi alive, not as a part of any 'enshitification' scheme.

Luis earlier described in multiple replies how AI usage is actually cheaper than normal search usage, as people tend to search less when they ask a question in the Assistant. However, your response here clearly implies that AI usage is more expensive than search usage. So which is it? If AI usage is in fact cheaper than normal search, then moving the Assistant to its own subscription (and seemingly not lowering the price of normal search) certainly feels like enshittification.

And even if AI usage is more expensive, the current model where Professional subscribers get access to cheaper models is a great selling point to help make the $10/mo price more palatable. The vast majority of people use little to no AI, and even when they do, cheaper models are more than performant enough for most questions; especially with Kagis search as its backend. Unless the cost AI usage is becoming untenable for the company, the bundled price of a good, ad-free search engine + good enough AI with privacy and choice of model is a great method of customer acquisition. Separating the pricing without a decrease in the price of the Professional plan changes the value calculus for your users, and not in a good way.

marc-eu

Vlad We can not afford to sell a product at loss knowing there will be fresh VC money in the next round, so who cares.

I would love to see Kagi introduce a token based search (not AI) service, just like f.e. MetaGer does. There I pay 1 euro cent per token (e.g. 5€ for 500 tokens), and every search costs 1.1/1.3 euro cent (there the costs depends on selected search indexes).

The advantage for MetaGer is that they never have to take losses, just because the user does more searches than the monthly subscription covers.

The advantage for the user is that they only pay for what they use, because the bought tokens are valid forever.

luked

I can’t speak for the existing economics Kagi has to endure for pricing, but one way of keeping most people happy might just be to draw a line in the sand, and grandfather existing plans in. That way Kagi is free to launch whatever new “clean” plans they like with economics that are more sustainable for them.

Give existing customers the ability to exercise renewing their current plan in perpetuity as a “thank you for helping us on our journey”, new customers go on the new sets of plans.

Eventually if Kagi’s growth is sustained the existing customers will be a rounding error.

Pleonasm

phunkytech ...cheaper models are more than performant enough for most questions; especially with Kagis search as its backend.

FYI: The assertion that Kagi Assistant is more performant because it is powered by Kagi Search may in fact be correct. However, as discussed here, there does not appear to be any objective evidence supporting this hypothesis. In fact, the Kagi Research Assistant itself says....

Kagi prominently claims that "AI models can perform up to 80% better when sourcing data using Kagi Search compared to other search engines". However, my research found no independent studies or peer-reviewed research that validates this specific claim. The statement appears to be a marketing assertion without publicly available supporting methodology or data.

SearchingAvocado

I don't use the assistant very much HOWEVER there are few cases when I do.

To me, this would be both a downgrade and not incentive to buy the standalone product. I used 0.22$ in 2 months so very rarely. But when I do need it I appreciate it for being there.

How about, let the pro users top up whatever amount they consider they need for the subscription period? Personally for me $5 would go a long way. I don't mind paying $5 extra but it also means that I don't have to subscribe just for assistant (which I won't anyway) and keep the same experience I have now 😉

lajawi

Reading through some of the comments in this thread, I feel like the odd one out not using the assistant or any AI at all. I've even gone so far as to hide it with CSS.

I'd personally love it if my current subscription (Professional) would lose the AI altogether, optionally with a decrease in price.

german

I subscribed to Professional today, because I wanted to leave Perplexity behind and I found no better option than Kagi. Throughout the trial, Assistant was generally enough to replace Perplexity for my needs. The fair usage model was attractive and was a large part of why I subscribed, because I was under the impression you have it all "worked out" and that it would prevent a rugpull, not create it.

Then, I saw this discussion and became extremely disappointed. Granted, this is partly on me for not doing enough research of your roadmap before clicking buy, but this still feels like a rugpull - right as I stand on the rug. A "one time" credit would really, really suck from the perspective of someone subscribing for Assistant, especially if they did so right now.

In the past I could not justify paying $10/mo for Search. Assistant is what brought me over the edge and made it justifiable.

Your marketing said "Kagi Assistant with Quick mode and a larger monthly usage allowance", not "Kagi Assistant with Quick mode and a larger monthly usage allowance, until we change our minds in a month and then this point is removed" - that really, genuinely, definitely doesn't read well.

markkrueg

I recommend never using the words “free” or “trial” as part of any paid-level service on Kagi. It just feels yucky. I would only use those words for the actual free tier that is completely unpaid.

For paid service I can see limiting a feature based on credits as how my Ultimate plan works, but any offer of free/trial is more likely to annoy the kinds of savvy users that pay for your services.

mixchiev

It would be great if kagi was a bit more clear about what the current issue is, because it is hard giving feedback without understanding the current constraints that make the current system unsustainable. I assume part of the problem is the various services they develop (eg translate) and how they are gonna be included in which subscription. Another issue, as I understand, is that the search-only people tend to be power users who use search very heavily, and a decrease in the pro subscription if separated from AI features is not very sustainable, but keeping the same price without AI features feels like a downgrade to different users.

Honestly if there is some issue that some of the pro kagi's services are not sustainable in a pro subscription, the only way I see forward is set some unified credit limit in all services the same way assistant has a credit limit. Similar for the ultimate subscription, and maybe also create another subscription in between. This way feels more uncomplicated to the majority of users that probably are not overusing resources, while letting them using the current services to responsible extents. If you don't want to include some newer services in the pro subscriptions, dont put them there and have them as add-on or sth. Nobody will be annoyed for not having extra services added, but people will understandably be if you break their workflows and then try to upsell them. But also nobody will reasonably be annoyed if you establish comprehensive general fair use limits, the way the assistant has.

Elkaki123

mixchiev

But also nobody will reasonably be annoyed if you establish comprehensive general fair use limits, the way the assistant has.

Im conflicted on this, on the one hand I like the idea because its fair, and as a bonus it would be good to understand how expensive search actually is, on the other hand, im a search only user (I dont even use translate), and I can see myself getting annoyed if I want to search something and I hit a limit at the end of the month, then having to get into duck duck go or something on the fly would get annoying fast and would probably end up switching search engines at that point. Paying for search was already unintuitive for me, though I became convinced it was worth it, but if with the current 10 dollars I cant use this as much as the free options I think Id inevitably end up leaving.

Thats my take on the matter of course, maybe I wouldnt hit the limit so who knows if its that bad of an approach.

Balkft6

Elkaki123 I agree with you on having no cap when searching (for the pro-subscription); a cap on number of searches would make me leave Kagi.com. I use the assistant and translator only for a limited amount (< 1 euro per month). A cap for free use of the assistant seems reasonable to me; if one exceeds the monthly payment with assistant/translator usage it is logical that one pays extra for that. Otherwise the other users will pay for it in their subscription price.
Probably users like us, with very limited usage of assistant/translator, are already paying "too much". But I am not bothered by that. I pay for road maintenance in my country and most probably partly for people that drive more than I do (I however visit the subsidized swimming pool more than the average inhabitant of my village).

Pleonasm

With all due respect to Kagi, one wonders whether the discussion of a standalone (Quick) Assistant subscription is still relevant, given that Google has now transitioned to an “intelligent search box” as a default? In this new reality, it is challenging to believe that such a product from Kagi could have compelling competitive differentiation and be commercially viable?

coffeefirst

Something that hasn’t come up — Assistant Pro with the cheap open models is what I need.

If you find a reasonable path at a fair price for:

Light usage
Small model usage

The actual sub structure isn’t that important to me.

coffeefirst

Something that hasn’t come up — Assistant Pro with the cheap open models is what I need.

If you find a reasonable path at a fair price for:

Light usage
Small model usage

The actual sub structure isn’t that important to me.

Pleonasm

marc-eu The advantage for the user is that they only pay for what they use....

Unfortunately, the interests of customers and the interests of Kagi are not aligned. When a customer makes fewer searches in a billing period than permitted by the plan, the forfeited searches (known as 'breakage' in marketing) become "100% profit" for Kagi. Therefore, it is not in Kagi's best interest to adopt a token based approach, even though doing so would be customer-centric.

marc-eu

Pleonasm Therefore, it is not in Kagi's best interest to adopt a token based approach, even though doing so would be customer-centric.

Well, I have an average of 700-800 searches per month, so according to some earlier post of Kagi (every search costs them 1.5 cent), they lose money on me. But still I would like them to do this, because I think that's the most honest approach.

NB: Just like paying per watch movie on Netflix.

Pleonasm

marc-eu I think that's the most honest approach.

Yes, I agree. However, I am not optimistic that Kagi will adopt a usage-based search pricing approach, because it would likely reduce revenue.

poinck

We are currently using Duo and it serves us well. If, for some reason, Search and Assistant become seperate subscriptions, I hope it won't become more expensive or it could add a lower tier option for both products (currently, we are not using our "ai cost value" to the full extent).

We like Duo, because one of us is more the Assistant user and the other one uses the search more. So, in effect, the current subscription model benefits us. For me, personally it would be ok, to have only an Assistant subscription as long as the Web search feature of the assistant utilizes all the benefits from Search.

« Previous Page