I have been using Kagi for around two years. If I don't find the results I want, I sometimes try Google and I don't find results there either.

Difficult to say if Kagi gives better results, but what's for certain is that it doesn't give worse results.

I am happy to pay 120$/year to support the business model.

I can't think of any other product that makes it to the HN frontpage so consistently. I guess Kagi really hits the bullseye in terms of HN as a target audience.

I just moved to Kagi last month. Here’s my experience:

Searching for real time information that happens within a day is way worse than google. I realise how much I missed the shopping search of Google. Google map is still irreplaceable. Google search has better UX, I particularly missed the favicon that identified the site next to URL.

Image search is also way inferior

I particularly like the Teams pricing model:

  - Professional ($10/user/month): Unlimited search
  - Ultimate ($25/user/month): Unlimited search plus LLMs

  You can mix and match users on different tiers.

Still, I'm only using Kagi Search and not any of their LLM Agent stuff.

I am still waiting for better pay-per-use LLM access, rather than $15-20/mo subs.

Nice! Happy user of Kagi for over two years here.

Suggestion: it would be cool if the Teams plan included a “personal” plan you can use with a personal account for each member. 1Password does it this way - if you have 1Password at work, you can “link” your personal account to it to get a family plan free of charge - but it’s linked for just this purpose, your employer doesn’t get access to your personal 1Password.

This is great to keep professional and personal accounts separate.

Funnily enough, deleting all of my raise/pin/lower/block filters dramatically improved the quality of my search results. Enjoying Kagi a lot more now.

I must be missing a minimum team size? Because if there isn’t one then why would anyone pay for “individual Ultimate” over “Team Ultimate” when they are the same price but you get more with the Team version?

> access to all premium LLMs in the world

That's a bold claim

What's the difference? Is it just being able to manage billing in a central place? The actual offering and price seem identical to just giving people their own account, right?

Two years with Kagi, sometimes go months without using Google. Stay pure.

Kagi plus a few good LLMs (I use local ones via ollama because I do that, and also Claude sometimes) gives you research capabilities that are as good as the pre-enshittification Internet and often better. I use LLMs like brainstorming partners and research assistants and then use Kagi to search for real data to verify results and make sure the LLMs aren't hallucinating.

Going back to Google I'm sometimes shocked at how absolute trash it is. I'm sure they're maximizing their internal KPIs.

Of course I know the real problem with Google is that I'm not the customer.

I've moved on from Kagi. Their service is absolutely better than Google, but not at least $120 / year better than Google. Especially when you have to deal with the plugin and private browsing nonsense that introduces a ton of friction.

Very interesting going after enterprise. How do you pitch to a CFO that you now need to pay for a web search license on top of Copilot/AI Assistant or whatever?

Would love to use kagi, but they (Google) block Mullvad from specifically the registration page.

Anonymity > decent search 8 days a week!

What's the difference between using Kagi and Perplexity? On X everyone talks about Perplexity, on HN, about Kagi. Do they both search -> put results in an LLM as input text?

Dropped Kagi after being a happy user and annoying all my friends about it for years. Sadly integration with Yandex is a deal breaker and a total violation of trust, sad.

Stopped paying for Kagi after finding out that they are using Yandex for some part of the search results and are paying Yandex 2% of their revenue for that.

For the unaware, Yandex Search is tightly integrated with the Kremlin.

I'm a Kagi subscriber, It is good, I am finding myself using perplexity.Ai quite about now it's very good.