Khoj is an open-source personal AI that searches and reasons across your notes, documents and the web — self-hostable anywhere.
Khoj indexes your documents, notes and PDFs, then answers questions, creates documents and runs scheduled automations using the language model of your choice. It can be self-hosted for full data control or used through the hosted service. Khoj also searches the web and blends results with your private knowledge. Its open-source nature makes it popular with privacy-conscious researchers and developers.
Khoj is free to self-host; the hosted service is freemium with paid plans for higher usage and managed convenience.
Privacy-focused researchers, developers and knowledge workers who want AI over their own data with self-host control.
Yes. Self-hosting is free and open-source; hosted plans have a free tier with paid upgrades.
Yes. Khoj exposes APIs and clients, and its open-source codebase can be extended freely.
Khoj’s self-hosting and model choice offer data sovereignty that cloud-only research assistants cannot.
Privacy-conscious researchers self-host Khoj to query personal notes and papers without sending data to cloud AI. Developers set up scheduled agents that summarize new documents daily. Writers search across journals and drafts conversationally. Teams deploy shared instances over internal documentation.
Self-hosting needs technical setup, and local model quality depends on hardware. Hosted tiers impose usage limits. AnythingLLM, private ChatGPT deployments and Notion AI offer alternative paths.
Khoj deploys via Docker or cloud hosting for self-hosters who connect their notes, files and model providers (local or hosted). Clients cover web, desktop and chat integrations. Setup involves indexing content sources and configuring model endpoints and privacy settings. Ongoing administration keeps indexes fresh and monitors resource usage for local models. Hosted plans remove infrastructure work for users who prefer managed service.
Khoj is the privacy-first choice: self-hostable AI over your own documents with model freedom. Setup effort is the trade-off, and local model quality depends on hardware. Developers and privacy-conscious researchers will value the control it provides over cloud-only assistants.