Kagi's Assistant is now available to all users, enhancing search with advanced LLMs

Kagi's Assistant is now available to all users, enhancing search with advanced LLMs

Kagi has expanded its Kagi Assistant feature to all users, which was previously only available to Ultimate plan subscribers. This feature is now accessible at no extra cost across all subscription plans and integrates advanced large language models (LLMs) from OpenAI, Anthropic, Google, and Mistral into the Kagi Search ecosystem. The Assistant supports tasks like research, summarization, coding, grammar reviews, and troubleshooting, complementing Kagi’s core search service while emphasizing responsible AI integration and user privacy.

Users can control the Assistant’s web access, choosing between grounded search results or direct model interactions, with support for personalized domain rankings, Kagi’s Lenses to refine scope, and file uploads for added context. They can build custom assistants with specific instructions and assign “bang” shortcuts for quick access via the browser search bar. Conversations are editable, allowing prompt adjustments, model switching, or setting changes without restarting.

Privacy is central—threads are private, auto-expire, and data isn’t used for training, while model usage follows a fair-use policy based on each plan, with most users staying within limits. Rollout is gradual, starting in the USA, and LLM access varies by plan, with fewer options on non-Ultimate tiers.

Our take: Mistral

by Mauricio B. Holguin

  • ...

Kagi Search is a web search engine that emphasizes a premium user experience by ensuring everything on the page is relevant. It is ad-free and does not track users, prioritizing privacy and a clean interface. Kagi Search includes features like Bangs for quick access to specific sites. It is rated 4.1 and commonly compared to alternatives such as DuckDuckGo, Brave Search, and Searx.

Comments

superstickynotemealt
1

Reading this I was like: Cool good excuse to start using them again, especially since their fair use model would work fine for me. I only use $3/mo in open AI API calls so the $10/mo plan is perfect, and then I saw the list of models for ALL vs Ultimate... what's the point of having fair usage linked to your monthly sub amount if you are going to make everything but ultimate near useless because of pointless model limits?

Gu