AnjadheAn AI assistant without internet is only half useful. Ask it about a price, a news story, or anything that happened recently, and it simply cannot answer.
Now web search is built into Anjadhe. No need to create any account or copy any API key. When you set up the app, it asks you one question: do you want web search? Say yes, and it works. Say no, and nothing ever goes to our server. 300 searches a month are included, free.
Why am I explaining so much about a simple search feature? Because search queries are very personal data. What you search about your health, your money, your family tells everything about you. I did not want to add search to a privacy-first app without thinking it through properly.
What happens to your search
When the assistant searches, your query goes from your Mac to Anjadhe Connect, a small server we run, which forwards it to a search provider and brings back the results. That is the whole trip. Four things we made sure of:
- We don't store your searches. Our server only counts how many searches you did. Never what you searched.
- We don't know who you are. Your Mac uses a random token. No account, no email, no name. There is nothing to attach a search history to.
- The search provider cannot profile you. They see searches coming from our one server, from all users mixed together. Not your IP, not your key.
- You don't have to take my word for it. The server's source code is public, so you can read it yourself. And in the app, Settings → AI Assistant → Web Search Logs shows every query that left your machine.
Being honest about it
When you search the web, the query has to leave your Mac. There is no magic that avoids that. That is what searching means. So the promise is simple and specific: only the search query leaves. Your documents, your emails, your notes, never.
And if you don't want our server in between at all, that is also fine. Put your own Tavily or Brave key in Settings, and searches go directly from your Mac to that provider.
This is how we will do any hosted feature in Anjadhe: say clearly what leaves your machine, keep the sensitive part out of our hands by design, and publish the code so you can check.