Khoj
Khoj is an AI personal assistant that serves as a desktop application for searching and interacting with your personal notes, documents, an...
Last verified:
What is Khoj?
Khoj is an open-source, personal AI app that serves as your AI second brain. It allows you to build custom agents, schedule automations, and research across your personal documents and the web. You can chat with any local or online LLM including models like llama3, qwen, gemma, mistral, gpt, claude, and gemini, turning any AI model into your personal assistant.
Key features include semantic search across your indexed notes and documents using natural language, getting answers from both the internet and your private files (including images, PDFs, Markdown, Org-mode, Word, and Notion pages), creating custom agents with tunable personality and tools, automating repetitive research tasks with personalized newsletters, generating images, talking out loud, and playing your messages. The app is accessible from web browsers, Obsidian, Emacs, the Khoj desktop app, phone, and WhatsApp.
Khoj is designed for individuals who want privacy-focused AI assistance with complete data ownership. It's ideal for researchers, knowledge workers, developers, and anyone who wants to organize notes, find relevant documents quickly, create automated workflows, and extend their capabilities with personal AI. The tool scales from on-device personal AI to cloud-scale enterprise AI.
The platform emphasizes transparency and adaptability, building tools you can understand that grow with you. It prioritizes privacy by allowing self-hosting on consumer hardware, ensures no model lock-in, and focuses on technology for human flourishing with intelligence you can trust.
Khoj pricing
Pricing model: Freemium
Khoj is free and open-source under AGPL-3.0 license. The cloud service (Khoj Cloud) was discontinued on April 15, 2026. Users must self-host Khoj on their own hardware for complete data ownership and privacy. There is no paid tier currently since the cloud service was sunset. Enterprise availability was mentioned as cloud service, on-premises, or hybrid solution before the cloud shutdown, but current pricing details for enterprise are not available on the website.
Khoj pros
- Open-source with AGPL-3.0 license
- Self-hostable on consumer hardware for privacy
- Supports any local or online LLM (GPT, Claude, Gemini, Llama, Qwen, Mistral)
- Semantic search across personal documents and notes
- Works with multiple file formats: PDF, Markdown, Org-mode, Word, Notion, images
- Accessible from Browser, Obsidian, Emacs, Desktop, Phone, WhatsApp
- Create custom agents with tunable personality and tools
- Schedule automations and recurring research tasks
- Get answers from both web and private docs
- No model lock-in
- Progressive Web App (PWA) support for phone
- Drag-and-drop document upload via web interface
- Incremental search-as-you-type experience
- Generate images within the app
- Voice functionality: talk out loud and play messages
- Experimental research mode with /research command
- Advanced natural language understanding using Transformer ML
Khoj cons
- Khoj Cloud service was sunset on April 15, 2026
- Requires desktop client to share and sync data with Khoj
- Self-hosting requires technical setup knowledge
- Document upload limited to one at a time on web interface
- Only one file upload per interaction on web
- No cloud service option anymore for non-technical users
- API rate-limited based on subscription tier (when cloud was available)
- GitHub repo shows 0 stars, 0 watchers, 0 forks indicating small community
- Tiny team of hackers may limit development speed
Frequently asked questions about Khoj
What is Khoj?
Khoj is an open-source, personal AI app that serves as your AI second brain. It allows you to chat with any LLM, get answers from your documents and the web, build custom agents, schedule automations, and perform semantic search across your notes using natural language.
How do I get started with Khoj?
Since Khoj Cloud was sunset on April 15, 2026, you need to self-host Khoj on your computer. Read the documentation at docs.khoj.dev for self-hosting instructions. You can install the Khoj Desktop app for your OS and configure it to connect to your self-hosted server.
What file formats does Khoj support?
Khoj understands PDF, plaintext, Markdown, Org-mode files, Word documents, Notion pages, and images. You can upload these documents to Khoj and it will index them for semantic search and contextual responses.
Can I use Khoj without self-hosting?
No, the Khoj Cloud service was discontinued on April 15, 2026. Khoj remains fully open-source, so you must self-host it on your own hardware for complete data ownership and privacy.
What LLMs can I use with Khoj?
You can chat with any local or online LLM including llama3, qwen, gemma, mistral, GPT, Claude, and Gemini. Khoj lets you turn any AI model into your personal assistant without model lock-in.
How do I access Khoj from different devices?
Khoj is accessible from your web browser, Obsidian, Emacs, the Khoj desktop app, phone, and WhatsApp. You can also install Khoj as a Progressive Web App on your phone for quick access from your home screen.
Can I create custom agents with Khoj?
Yes, anyone can create custom agents with tunable personality, tools, and knowledge bases. You can create agents to take on any role with custom knowledge, persona, chat model, and tools.
What is the /research command in Khoj?
Starting any message with /research tries out the experimental research mode with Khoj. This mode is designed for deeper research tasks and iterative discovery across your notes and the web.
How does semantic search work in Khoj?
Khoj uses advanced natural language understanding with Transformer-based ML models for semantic search. It provides incremental search with a fast, search-as-you-type experience to quickly find relevant notes and documents using natural language.
Is Khoj free to use?
Yes, Khoj is completely free and open-source under the AGPL-3.0 license. Since the cloud service was discontinued, you self-host it on your own hardware, which means no subscription fees, though you'll need your own hardware and possibly your own LLM API keys.