Provide an API for the Assistant

pbronez

mateja That sounds good to me. I like the clear delineation of the "extra" bucket.

FWIW, I've seen several of my AI-focused vendors adopt a Seat + Usage model like this. It's complex, but done well I think it can align incentives correctly.

The upfront per-seat cost needs to cover the vendor's fixed costs to support my account and make a profit without being so expensive that I start scrutinizing who gets a license. The main way vendors can reduce this upfront cost is reducing the included usage.
The Usage cost needs to have good margins for the vendor so they make more money when the product is good and we use it more
I want options to reduce costs if I wind up with consistently high usage. These can include bulk discounts, pre-pay discounts, etc. They should reduce but not eliminate the vendor's margin.

That's all very enterprise-y. From a personal perspective, I love my Ultimate Account, and I would love it more if it cost less. I don't expect "Unlimited" usage. I'd prefer a lower subscription cost with the extra bucket.

TomatootamoT

Is there any idea on if there is going to be a rate limit? Or will there be different limits based of plan type?

pineapplemachine

I'm also interested in this, especially since the Kagi-specific Quick and Research assistants came out and they do not seem to have any existing API that I could use. I'm not that particular about pricing or billing. It would be nice if API usage would first go through the token allowance already included with my Ultimate plan, but that's not an absolute requirement for me. I have appreciated the usage-based billing for FastGPT, though, meaning that while my API usage is low, so is the pricing.

Pasithea0

I'd love to use Kagi Assistant within Obsidian!

person085696

Suggestion if on iOS, considering Vlad said the API isn’t likely to happen:

Create a web app to essentially emulate a native app. Best convenience and functionality under these specific circumstances.

jakar

ncwade

Would love API access, more than happy to pay for all usage via the API. Gives me A) a single subscription, and B) a source I trust to sit in front of the companies I don’t.

I want to second the sentiment behind this. Having a trusted party like Kagi eliminate some of the more annoying aspects (like multiple subscriptions / API keys, keeping up with all of the tradeoffs of a constantly growing and changing set of models, etc) has solid value. And I could be wrong here, but I feel that Kagi users, at any level, are a demographic that understands, better than most, how this all needs to be economically sustainable to be useful.

westerntragopan

jakar I agree, this is one subscription I'd happily pay for for the rest of my life. For someone who's planning to study abroad in Europe, where the country's first language is not English, Kagi especially Translate is going to be such a lifesaver for me.

kernst

jakar Agree on all counts.

In addition to the already-mentioned uses cases, I'll add one more: Firefox's recently added chatbot integration. It'd be lovely to be able to pair that with a Kagi subscription, via an OpenAI-compatible API endpoint.

Why would Kagi do this, or why would I pay for this? In a word: taste. I trust Kagi to choose an affordable general-purpose LLM for me — omakase-style — in the same way Kagi Assistant has a default model. All I'd be looking for is something that works reasonably well for asking questions about the currently-viewed page, using the native UI of my preferred browser.

Yes, Orion, I know, I know. If anyone thinks this is a real problem to "help" another browser vendor, then they just need to picture the pie getting bigger, instead of Kagi's piece getting smaller. :)

Disregard! See octogiraffe's reply, below. I momentarily forgot that Firefox's "AI Chatbot" is just a sidebar that loads a web site, and stuffs some text it in. It's a small wonder that this actually works, but it totally does, even with a llamafile.

octogiraffe

kernst

kernst All I'd be looking for is something that works reasonably well for asking questions about the currently-viewed page, using the native UI of my preferred browser.

FYI this is already possible! I don't use it much but it was easy to set up. I haven't had much luck with theming it, but thankfully that isn't watcha need

Just set browser.ml.chat.provider to https://kagi.com/assistant in your about:config, works like a charm. I'm personally wary of any sort of AI integration into the browser, but it does work, even if the prompts "look" a little rough around the edges

Sorry that this is offtopic to the thread but I couldn't help but notice that this is already possible <.<

RyanPrice

I just wanted to voice my support for the addition of API access to the assistant. I have so many use-cases I would love to utilize this for.

sTiKyt

strategic040 i don't know where are you getting that from 3000$ to 20000$, maybe in US it is, in the rest of the world IT goes as low as 300-500$ a month, many are happy to make at least 1500$

jakar

sTiKyt yeah, those are crazy numbers, at least at the high end. $20k/month is $2.4M/year. And I think they're talking about net income (since it's monthly), so that $2.4M net would be well over $3M gross annually. There might be someone making that much somewhere, but it's very very far from the median IT professional or engineer. If I'm wrong, then I need one hell of a raise, I guess.

Phil

I would love to have a pay as you go API with a slight markup over the providers because Kagi has a better privacy policy than OpenRouter or EdenAI, companies that provide this sort of service.

I don't even need kagi search as part of the assistant API, I can use it via an MCP server or similar myself.

oxlvlnle

Phil But since Kagi assistant is backed by these services, the difference is negligible here...If and only if Kagi host their own LLM.

Drizzt321

Not having read through all of these posts, but my ideal would be something compatible with Ollama API, or accessible as a backend via Ollama frontend. My use case would be Home Assistant Voice Assistant with Kagi Quick for basic searches/lookup that is quick. Of course I'll then have to figure out how to fallback (if not understood) to a slower, local (via GPU) Ollama LLM with Tools (e.g. control Home Assistant devices).

trupi

Did you consider wrapping Kagi in a CLI? That could be a middle ground between fully opening the API, but still providing some way to programmatically interact with the service. Even if you'd expose only Quick and Research with similar experience of the Web, it would basically add an ability to pipe the response out to some other tooling. With the CLIs being all the rage recently and tech moving really fast in that direction (using web tech stack in terminal) maybe it'd be easier for you from technical perspective?

This is what Obsidian has done recently: https://help.obsidian.md/cli

I love Assistant; it's a real lifesaver. But about 20% of my usage is asking Claude Code to write a research prompt I can manually copy into Kagi and then paste back the response from Research agent. In that scenario the question is not mine, and the response is not to me. Both CLI and API would make it way easier.

westhehacker

You know, I have been wondering about how Kagi makes ends meet every month with how much LLM we get access to for only $25. Now that I know that y'all are being exceptionally generous with token limits, I'll try to cut back on my LLM usage as much as possible. Maybe I'll cut down to almost zero and just let you take the extra $15 per month as a donation. I'd be down for lowering the token limit every month to just $5 or $10. I really like Kagi and want to see it prosper. I've hit the monthly allowance a couple times before and just renewed my plan early so I could use more because I thought you were still net-positive with it. Even if the price goes way up or my allowance drops a lot I totally support whatever is best for the future of Kagi, and I would like to see API access available for the Assistant, perhaps completely separate from the monthly client allowance. I'd imagine you'd get $5 or $10 to use in the Assistant client from the Ultimate plan and then pay for token credits that are for API access only, the two shouldn't be mixed. I think that makes the most sense. I'm an electrical engineer and having fast, reliable information synthesis, source colleciton, and data analysis is invaluable for pushing complicated projects forward.

westhehacker

I definitely misunderstood the token limit, I thought you were making more money by me hitting the max and renewing early, sorry about that hah

jankais3r

Basically everyone in this thread is an advanced user asking for API access not to save money, but to get access to various models through a trusted provider. I don't really understand the economic argument against offering this as a service. Kagi can simply charge us per token use, add a little margin on top of what they pay to their upstream provider, and everyone will be happy...

pbronez

My opinion on this is shifting a bit. I set up an OpenClaw powered by Haiku via OpenRouter and blew through $50 in a single day. That was expensive enough for me to turn it off. This makes me think I want my LLM API access separate from my personal/direct AI access. I'd hate to lose access to Kagi Assistant in my browsers because an AI agent got to greedy and blew out my budget for the month.

Honestly, if the economics are the issue just roll it into the existing API platform. Buy credits, use credits, exhaust credits, buy credits. Existing Kagi subscription is just a prerequisite for access to the API. Totally separate usage pools.

I'm going to play around with OpenClaw powered by a local LLM with access to Kagi's search and FastGPT APIs. Might strike a nice balance.

Vlad

pbronez Yep this is how we are thinking about it. API access separate from UI use. We are working on this API currently.

« Previous Page Next Page »