Khoj is the open source personal AI assistant designed for deep research: it indexes your Obsidian notes, Org files, PDFs, and Notion pages alongside live web search to answer questions with citations drawn from your own knowledge base.
The Problem
Perplexity and Notion AI are cloud-only. Every document you connect and every question you ask is processed on their servers. For researchers, knowledge workers, or anyone whose notes contain sensitive professional information, this is a meaningful privacy trade-off. There is also no way to make these tools aware of private documents without uploading them to an external service.
How Khoj Solves It
Khoj runs on your machine. It embeds your documents locally and stores the index on your own disk. When you ask a question, it retrieves relevant chunks from your personal corpus and the open web, synthesizes a grounded answer, and cites its sources. Custom agents let you configure specialized assistants for recurring research tasks; scheduled automations run research queries and deliver summaries on a timer.
Key Features
- Indexes Obsidian, Org-mode, PDF, Markdown, Notion, and plain text files
- Hybrid retrieval: searches your local knowledge base and the open web in one query
- Custom agents for specialized workflows (research, drafting, summarization)
- Scheduled automations for recurring research queries and digest delivery
- Multi-modal: image understanding and voice input
- Self-hosted with fully local embedding and retrieval
Who It's For
Khoj is best for researchers and knowledge workers who take detailed notes and want AI that knows their documents, privacy-first users who cannot upload work materials to cloud AI tools, and teams running internal knowledge base search without a SaaS subscription.
Compared to Perplexity
Unlike Perplexity, Khoj runs entirely on your own infrastructure — it indexes your private documents locally and combines them with web search to produce grounded answers, without sending your notes or queries to any external server.
