Khoj

Unlike cloud-only AI assistants, Khoj can be run entirely on local infrastructure, giving privacy-conscious users and organizations full control over where their data resides and which language models process their queries.

The platform works with any online or local LLM including GPT, Claude, Gemini, Llama, Qwen, and Mistral.

The document retrieval capability allows Khoj to answer questions using personal knowledge bases built from image files, PDFs, markdown documents, org-mode files, Microsoft Word documents, Notion pages, and other formats.

Users ask natural language questions and receive answers grounded in their own documents rather than generic web results. This makes Khoj useful for individuals who want AI assistance that reflects their specific notes, research, and knowledge rather than general internet content.

Khoj's custom agent builder allows users to create specialized AI agents with defined knowledge bases, personas, chat models, and tool access.

Each agent can be configured for a specific role, such as a research agent with access to academic papers, a coding agent with access to a codebase, or a support agent trained on product documentation. The custom agent framework extends Khoj from a personal assistant into a platform for building purpose-built AI tools.

The platform supports scheduled automations that run research and information retrieval tasks on a defined schedule without manual triggering. This is useful for recurring research workflows, monitoring tasks, and information aggregation that a user needs regularly but does not want to manually initiate.