This is quite specific but feel like this is the crowd for it!
I've been using perplexity for a while and really like it but would prefer kagi in anticipation of any enshittification.
Has anyone managed to make the kagi assistant more similar to perplexity?
Specifically what I liked was not drawing from the models knowledge but always from sources so I can go deeper and/or check the answers. I'm hoping a bit of prompt engineering should get me there! But quite new to it
you're assuming that kagi doesn't enshittify :)
Also true!
i'm guessing, these days, you'd want a model to be run with support for MCP so you can hook it up to a "search results" MCP or even a "read this web page" MCP
https://mcp.so/
does kagi support this?
https://github.com/kagisearch/kagimcp
Looks like it's in closed beta
I definitely use Kagi less since I started using Perplexity, enough to drop down to the Starter plan.
It's a shame that their AI product isn't as good as using Perplexity on a free plan. Feels like a missed opportunity for them.
Maybe I'm misreading https://www.perplexity.ai/hub/technical-faq/how-does-perplexity-follow-robots-txt , but it seems like Perplexity still isn't respecting robots.txt
entirely.
Yeah, Perplexity started out less ethical than even OpenAI, so it doesn't surprise me that they're still so problematic
Yeah this is my concern with them and why I want to make kagi work. They don't seem like they are in it for the good vibes!
I've tried adding a prompt to encourage references which works some of the time! Need to keep tweaking I think.
well ... i'm not sure kagi passes the vibe check either ( https://d-shoot.net/kagi.html ) but we all have different values and priorities and tolerances :shrug:
Last updated: Jun 28 2025 at 13:25 UTC