Kagi Assistant standalone subscription

onthefence

Vin $10 a month for a pro account is $15 CAD, and there is no way I would ever pay $25 USD. I am one of the outliers in that I am looking again at Kagi to see if anything has changed over the months since I last cancelled my subscription for Pro. I took out the subscription because of Assistant rather than search.

I found that I resorted more to free Gemini, ChatGPT Claude and paid Perplexity.

nzo

I quite agree with @Breulkf. I subscribed to the Professional plan with Assistant included, removing it would mean simply downgrading my plan without lowering the price.
Envisaging to reduce tokens/month maybe ?

Luis We know there are Starter and Professional subscribers who find utility in the Kagi Assistant but haven't upgraded to Ultimate, and we'd love to understand why. What would make the upgrade worth it for you? If you're using another tool, what's missing from Kagi Assistant to justify you going through the pain of switching?

I'm using principaly Le Chat from Mistral as conversational agent, so moving to Ultimate plan could be interesting (more models from different compagnies and a more attractive price, only a 5$ difference for the whole Kagi ecosystem), but not in the actual state. Compared to Le Chat, ChatGPT, Gemini, Claude, etc there is a long way to go, it missing features like Projects, Library, extensions, etc. For a similar service like Assistant, I can go for mammouth.ai which is cheaper and more sovereign for EU subscribers. If It came close to other conversational agent and allowing API access (I don't know if this is already the case), I could consider the Ultimate plan. On the other hand, the taxes will certainly question myself :/

jlharter

I quite like the AI assistant for some run-of-the-mill questions and its summaries. One reason is I block Reddit (too distracting!) with NextDNS. But often search results for questions point, rightly, to Reddit. I use the AI Assistant to quickly summarize those threads in the search page. Saves me from having to go disable my filter, and, frankly, it's easier than thumbing through all the comments.

Likewise for any search that has mostly obvious answers. Earlier I wanted to know, "Are there any workarounds to closing tabs remotely with Brave?" I didn't think so. But I don't need or want to thumb through endless listicles about Brave's sync chain function. I thought it was no, it said no, I moved on.

Removing this functionality would be a reduction in my existing plan, and, subsequently, a price increase on "less".

I am sympathetic to people who do not want any AI-anything in their feature sets. For those people, there should be a single nuclear option to toggle it off (Firefox is handling this pretty well.) But for this narrow case, I like the AI features because I trust Kagi more than I do ChatGPT, Claude, or any other model (certainly not Gemini). I suspect and trust that when I do a Kagi search, Kagi summary, or Kagi AI-anything, that the only people that know about it are me, my browser, and a temporary chat session with the model.

PHLAK

I rarely use Assistant and wouldn't really miss it if it was removed from the plan, especially if that meant a reduced price for a search only plan. That being said, it was a nice to have on occasion and removing it from the plan feels like a downgrade. I already feel that $10/month is on the high side for search only but I started paying for it before Assist was added so I'm not going to complain too much. However, having a small monthly budget for AI use would be nice. My highest month usage was $0.037 so I personally wouldn't need much.

dreifach

Luis What will Assistant access look like within family plans, when the individual user is not upgraded to Ultimate?

Alfondorion

A ~$1/month budget on all models for the pro plan could be a nice gesture

Breulkf

Alfondorion Frankly. I'd take a 90% reduction as an offense. Take away some of the less used models, for instance. This I could accept, but changing an in-place agreement is really a hard pill to swallow. To be frank, any change/reduction to my actual plan would me me cancelling it instantly.

Pleonasm

How might existing customers with the Professional Plan react to the proposed sequence of events, from an emotional perspective?

Step 1: Kagi removes Standard AI capabilities from the existing Professional Plan, with no ‘grandfather’ provision, and with no subscription price reduction. Customers become disappointed.
Step 2: Kagi requires those same individuals who want at least some AI capabilities (as were previously included) to purchase the standalone Kagi Assistant in addition to their revised Professional Plan (i.e., the Ultimate Plan ‘bundle’). Customers become irritated.
Step 3: The monthly cost increases by a factor of 2.5, from $10 to $25. Customers become angry.

It’s challenging to envision a scenario where this sequence of events will help Kagi build stronger customer relationships.

buster

As a heavy user of kagi and Assistent I think the current model is nice. I am an ultimate subscriber (so nothing changes for me, right?), but I was a pro subscriber and that led me to ultimate. I most likely wouldn't have upgraded to ultimate without Assistent in pro.

Last but not least, kagi gives me a good all around privacy minded feature set and I do think Assistent is and was a great thing to do, despite all the blame.

Gredharm

buster That's how I ended up on Ultimate. I only wanted Pro at first, as a replacement for Perplexity at the time. So I was using Assistant with GLM and Deepseek and I found it pretty good, but just not as good as Claude Sonnet or Grok, so I moved to Ultimate.

manchmalscott

All I want is a subscription that explicitly does NOT include any AI features. Not a trial, not a reduced usage cap, nothing. I want to pay for search. I refuse to pay for AI that I don’t want and won’t use.

I stopped being a paid Kagi subscriber years ago. A tier where I don’t have to see a single AI related “”feature”” is the only thing that would bring me back.

JakeKlingler

I fail to see how any current Pro subscriber could view this positively. You would objectively be paying the same amount of money for less capability.

As a current Pro subscriber (since 2021) and a very light user of Assistant, I can confidently say you will net less money from me as a customer if this change gets implemented. I do not get enough value from Assistant to warrant a standalone subscription, and would likely downgrade my search subscription to Starter or cancel it altogether in light of alternatives.

Gredharm

If the problem is simply that the current price is untenable due to the cost of search already being high, plus AI adding to it, then I think the solution is to just say that and increase the price. Maybe let existing users keep the current price for 6 months or a year and new users have the new price.

It's better just to own it and say, "Yeah, we need to increase prices." than to juggle around features to try to cut costs IMO. I mean 4 years ago, I was paying less than 3 Euro for a latte, now I'm paying 5, the world's gone to hell, we all know it.

grkgwkh

The way this was communicated is quite unfortunate. The initial post was pretty vague, later clarifications seem contradictory (Assistant trial only <-> Assistant would always be available to some extent), and now there's addendums all over the place in a 100-post thread that one has to read to understand that… it's all unclear.

What's left is confusion, a bit of a loss of trust, frankly, and my understanding that roughly you're asking “We'll probably be taking away features and you'll be paying the same, but we don't really know what we'll do – what's your feedback?”. What I'm also taking away from this thread is ideas from other people for how I might replace Kagi, sadly.

As for what the change would mean for me: I came to Kagi for search, back when Assistant did not even exist. I loved the search and I was very pleasantly surprised when Assistant was added. I convinced three family members to join, partially because of Assistant. Then search deteriorated massively for me (I posted about it here many times, no solution in sight), so Assistant has pretty much become the main selling point to me.

With Assistant becoming its own subscription on top, I would most probably leave and I would not expect my other family members to really care enough to set up their own subscription so bottom line: it would probably mean 4 customers or one family subscription less. I'm not saying this out of spite, it's what I would expect to happen.

rhyxalia

I'll add my 2 cents, but first some background for context. I'm a Kagi Early Adopter, on the annual Pro plan from the very beginning. I've also paid the annual price for two seats of ChatGPT Pro for 2 years now, and I use ChatGPT extensively for work in the IT field as well as to enrich my life. A few months ago I decided to cancel my upcoming ChatGPT annual renewal to switch to Kagi Assistant.

Reading through this thread, there are too many people here who threaten to jump ship at a drop of a hat. People who are comparing Kagi Assistant to one LLM subscription or full on AI Agent Orchestration. People who are wanting changes done to their benefit without thinking of how it would affect others. And those who are hardline anti-AI. Those groups do not care about Kagi or the benefits brought by Kagi's privacy and customization.

For those of us who are not in the groups mentioned above, many of us do care. Kagi has earned our trust so far. I would rather pay for the Ultimate subscription or for the additional Kagi Assistant subscription, than to leave Kagi to a worse experience because of a few bucks. Or leave because Kagi isn't my definition of perfect. I am paying for Kagi not only to show my support to a good organization, but also because I can see the value of things they are offering which are worth more than just price.

Yes, taking away or limiting Kagi Assistant is a shame, if that is what ends up happening. But at least Kagi is communicating with us very early on, and not just asking for our feedback, but also replying and actually having a conversation. Will Google or Anthropic be doing that?

howie

Kagi thinking that Assistant is worth more money given its long history of completely destroying your context files, summarising your requests, and in every way possible generally getting between you and the LLM APIs you're supposedly paying for completely non-transparently is a little funny to me. Assistant is being positioned as a premium product here when it's actually quite the opposite - it is simply the least functional LLM chat wrapper that I know of, due to deliberate design choices that the team stood by for over a year at least and didn't at all understand the impact of.

I have heard this has actually been changed, but this isn't the first time I've heard that it's been "improved" and it was never improved the last times and the team doesn't really seem like they have really grappled with the fundamental breakdown of trust associated with being the platform actively putting massive barriers between you and the LLMs without ever informing you - there's never been an understanding that this means I can never trust it. For a 20% markup on tokens! Do I really want to re-sub to Ultimate and actually try going through all of the multiple frustrations I've had every time I try again when I at least know Claude's direct subscription isn't cutting my uploaded files in half and serving markdown summaries of HTML code?

Assistant would need to both be a significantly better product and actually prove that it's going to remain a significantly better product before I can take this seriously. Kagi seriously need to rebuild my trust in Assistant before they claim for a second that it's premium enough to justify its own subscription. "Assistant has been unusable for a year in a way anyone could verify immediately but don't worry guys! It's good enough to be worth paying for separately now!" is not serious.

On another subject - I like using LLMs. I'm happy to use a good LLM interface. However, people have been asking for a way to pay for Kagi without the assistant and LLM features and have been told repeatedly that there's no point because LLMs are actually the cheap part of the subscription for them that they make money on. How can they square that with this? It seems like the viewpoint has completely changed on this - and worse, people will be getting their AI-free Kagi subscription at the same price. may not affect me as a non-assistant user but right now, this will ultimately be people paying the same amount of money for less; it is strictly a value-loss.

Ultimately, maybe this is how the business has to survive - if costs have gotten away from Kagi, of course they'll have to do something, either raise price or reduce value, and this probably beats raising the price. However, the communication has been confusing, and the motive still hasn't truly been explained, even if it's clearly for this reason. Euphemistically calling a direct cut to what we get for our money an "evolution" of pricing is quite literally what San-Francisco based tech companies say in emails that attempt to justify how service cuts for your money are actually a good thing, and you're supposed to be better than that - I find the doublespeak more offensive than the change itself.

Pleonasm

howie Assistant would need to both be a significantly better product and actually prove that it's going to remain a significantly better product before I can take this seriously.

In my opinion, this is a point that Kagi would do well to seriously ponder: is Kagi Assistant truly a better AI research solution than the competition - and, if so, will it remain competitive? Regardless of the current perceived merits of Kagi Assistant, it does seem that the competition is rapidly advancing while Kagi is remaining relatively static. For a company that seeks to differentiate itself based on product features, this evolution may not bode well for the future of the firm - or, for the launch of a standalone Kagi Assistant subscription.

The good news is that Kagi still has time to place a renewed emphasis on the development of Kagi Assistant. Doing so is likely to be more beneficial than modifying the subscription plans, which is akin to "rearranging the deck chairs on the Titanic." Updated subscription plans provide the appearance of meaningful change, without altering the underlying substance.

joshsmoshgosh

Luis are you able to provide stats on how may subscribers will be affected by the change?

I feel like I get a lot of value out of the assistant on the Pro plan and also Kagi Search. I don’t think I’m hitting the current cap, either. Probably won’t hit it by the end of the billing cycle. I’d be bummed if I was suddenly getting less.

joshsmoshgosh

joshsmoshgosh
I wanted to clarify what I use in the Pro plan. I subscribed for a year and I get a lot of value out The Assistant. Kimi K2.5 meets all of my search and code needs.

When I'm not finding what I need fast enough with a regular kagi.com search, I'll jump to the assistant, phrase my search as a conversational question, and go from there. I treat the Assistant as search engine most of the time. And I use it to start personal coding projects and look up libraries and basic use. It's this real nice sweet spot between search engine and coding agent.

I don't really touch Kagi Translate, but I'm also an average white dude living in the US suburbs. I don't need to translate much. Just not a part of my regular needs.

I've dabbled with the kagi search MCP server, but it also seems cost-prohibitive in terms of API credits and that it needs to use the summarizer makes it a non-starter for me, especially when I have unlimited searches on the website already.

So yeah, if I suddenly lost access, or had much more limited access to the assistant where I had to worry about tokens and credits, that'd make me rethink my plan when renewal time comes. I don't need access to a dozen models. Kimi K2.5 is good enough.

Is the number of available models a perk for others?

t8mdjrak

Luis Hi, I think there are some issues with the current business model with Search + Assistant so I'm glad there's a discussion thread here. However, my reason is quite different to Kagi's.

If I'm not wrong, Kagi is trying to find a way to attract more users to Ultimate plan? I honestly don't think this plan could work well enough for you:

As others said, $25 is too much. They are only potential for users who want to use AI a lot as it could help increase their personal income greatly. However, these users mostly are more attracted to some exact AI services who release better models regularly, have more features and more flexible ways to integrate to users' own devices/softwares (via APIs, configs...), rather than having ability to using more models. As time goes by, features + integrations would be more valuable than having more AI models. These users would be quite ready to jump the ship to other AI services if they see a better one for their work.
On the other hand, users who use AI occasionally definitely won't pay for $15 more, but as you can see others' replies, it's another nice feature to justify our current $10 subscriptions (which is already on the edge of expensive sides that many of us did consider more than once not continuing with Kagi any more) and it's always there when we need it. If it's completely gone without upgrading to Ultimate plan in a foreseeable future, many would feel it's a let down and it would just push their decision of not continuing the $10 subscription more.
Also, users who haven't used Ultimate plan before have no ways to know how better the more premium AI models would be for them and whether they could justify it as a worthy additional feature for $15 more. So naturally even when they bite the bullet to subscribe for a short time, they would have much more expectations and probably don't have enough time to really evaluate the more premium models.

In general, I don't think this Ultimate plan could really scale to sustainable point on my view. For me personally I would like to see a business model where:

Free plan as currently,
Pro plan that could access to all models (that would encourage users to use your Assistant models more, especially the more premium ones). Kagi can set the monthly cap limit to whatever you think it reasonable and sustainable for you ($1, $3, $5, or current $10, any is fine),
A credit system for Assitant alone where users could tip their money to that and the Assistant usage token cost would be taken from that credit when the monthly cap limit in point 2 is reached at that month.

bvsveera

My thoughts. As someone who is currently subscribed to the Pro tier ... this feels like a rug pull. Pro users would gain nothing in return, and still be expected to pay the same price, despite hearing multiple times that search + AI users cost less to serve than search-only users.

Prior to Kagi, I never used AI beyond the limited ChatGPT integration in Apple Intelligence, because I never want who I am to be associated with what I search for. That's exactly what happened with Google search, and I fear is what is going to happen if you use a model from the maker themselves (see ChatGPT's ad insertion). In the same way that Search's proposition is ridiculously good and private search, Assistant's proposition should be ridiculously good and private AI. As far as my use cases are concerned, that is how it stands today, and I have been a happy subscriber thus far.

I agree that there needs to be differentiation between the Starter, Pro, and Ultimate tiers, but I don't think this is the way to go. My suggestion would be to limit the Starter tier to Kagi Quick (or search-only if there must be a tier with minimal/no AI), Pro to standard models, and Ultimate to all models. In fact, I would like Pro to have access to all models as well, and if that means hitting the $10 monthly credit, then so be it. If providing that many credits is the real issue at hand, I would even be happy to have it halved to $5 credit. I find Assistant to be incredibly useful almost every day, but I only ever use $1-2 worth of credits per month. For that reason, I cannot justify upgrading to Ultimate. AI is useful, no doubt, but not worth an extra $15 USD per month for me. And, from what I can gather from the power users in this thread, Ultimate does not offer enough to justify a separate subscription. Not to mention the lack of mobile and desktop apps, nor API access.

I hope Kagi reconsiders this direction, and provides more concrete information in the future. At the very least, I thank you for making us aware of these potential changes beforehand.

all-good-usernames-are-taken

I would still want access to Quick Reply within the Pro tier of Search.

MattIvan

Not a fan of the idea. I barely use assistant as it is (don't really like the interface, I already have perplexity etc), but don't like the negative change and it feels wrong. I do use it from time to time and it enhances the value of my kagi subscription. I've been very annoyed at the enshittification of perplexity lately and have been evaluating assistant.

I think it's fine if assistant is spun into a separate subscription, and I think that's a natural product line and makes sense for you to do, but the idea of changing full, unlimited assistant to a temporary trial thing for AI on the professional plan is a hard no from me - you're making the product I already have worse if you do that and that... is not what I'm paying for? Isn't that against the kagi ethos?

Why don't you just implement a reasonable usage limit for assistant on the professional plan instead? - I think that would be ok. Something sustainable but also not 0. It should work for most users but keep costs low enough to be sustainable.

The professional plan could have unlimited traditional search + some reasonable allocation of assistant usage (across all models please) and then if you want more AI you can pay more for it. AI isn't free. I understand.

Alternatively, there's another path: grandfather existing plans as they are and subject new subscribers to the changes. It creates a strong lock-in effect for existing users. This is likely more expensive than the 'just have some reasonable limit' kinda thing but would be the most popular option.

MattIvan

There's a general problem with kagi pricing. It goes 5, 10, ... 25.

Have you considered a $15 tier? That's a lot more likely to work.

« Previous Page Next Page »