Khoj

Khoj AI's key features include the ability to chat with any local or cloud LLM, supporting fully offline or internet-based AI, and to retrieve and semantically search answers from web resources and your private documents. Users can create and customize AI agents with specific personas, knowledge bases, chat models, and tools to automate research tasks and workflows. Khoj can be accessed anywhere via browsers, desktop apps, mobile phones, Obsidian, Emacs, WhatsApp, and integrates with note-taking systems like Obsidian. It offers multimodal capabilities, allowing it to understand complex documents, generate images, diagrams, charts, and support voice chat. Khoj is privacy-focused, ensuring that when running locally, your vault data does not leave your device, with only AI models initially downloaded externally. Additionally, it provides automation and notifications through personal newsletters, smart alerts, and automated research task management.

Yes, Khoj can be used immediately via their cloud app at https://app.khoj.dev without any setup.

Khoj supports PDFs, Markdown, org-mode, Notion pages, Word documents, plaintext, images, and more.

Users can build customized agents tailored to their needs with specific knowledge bases and personalities. Guides and community support are available on their Discord and documentation.

When self-hosted locally, user data (e.g., Obsidian vaults) never leaves the machine. Only AI models are downloaded from trusted sources.

Khoj can be self-hosted using simple commands, with environment variables for API keys (OpenAI, Gemini, Anthropic, etc.) and admin credentials. It also supports one-click deployment on cloud platforms.

It runs on desktop, web browsers, mobile phones, and integrates with popular tools like Obsidian, Emacs, WhatsApp.

Users can maintain multiple chat sessions organized by topics, enabling easy continuation of discussions.