Self-Hosted / Projects / Khoj
Khoj
StableSpecs verified September 24, 2026 · against vlatest
Overview
Self-hosted personal AI for chat, semantic search, research, and document-aware agents.
Khoj is a self-hosted personal AI and second brain for searching notes, documents, and the public web with natural-language questions. It is aimed at people who want a private assistant grounded in their own information rather than a generic chatbot with no access to their files.
Why people choose Khoj
Khoj combines semantic search, document-aware chat, web research, and personal automation in one application. The official docs emphasize local search and indexing for self-hosted deployments, while also allowing users to connect local or online language models depending on their privacy and quality requirements.
What is included
The documentation lists support for PDF, plaintext, Markdown, Org-mode, and Notion content, with access through a web browser, desktop app, Emacs, and Obsidian. Khoj also documents custom agents, scheduled automations, image and voice capabilities, and research workflows, though requirements vary by feature and model.
Requirements and tradeoffs
Self-hosted Khoj is more than a static file viewer: indexing, embeddings, model inference, and optional media features consume resources. You can connect a local model for greater control or an online provider for easier setup and potentially stronger results, but hosted providers change the privacy boundary. Follow the current setup guide for supported Docker and model configuration.
Best fit
Choose Khoj for a private personal knowledge assistant that can search your notes and documents across several clients. If you only need a lightweight full-text search engine, a dedicated search tool is simpler; if you need a team-wide observability or workflow platform, Khoj is the wrong category.
Interface previews
Screenshots of the Khoj interface.
Feature support
| Feature | Support |
|---|---|
| Natural-language document searchThe official docs describe semantic search for relevant notes and documents. | Supported |
| Chat grounded in personal filesKhoj can chat about shared files and use relevant documents in responses. | Supported |
| Web research and web answersThe docs describe access to information from the public internet and research workflows. | Supported |
| PDF, Markdown, text, and Org-mode ingestionThese formats are listed in the official introduction and data-source documentation. | Supported |
| Notion integrationThe official docs include a Notion data-source integration. | Supported |
| Web, desktop, Emacs, and Obsidian clientsKhoj documents each of these access paths. | Supported |
| Local and hosted LLM providersThe project supports local or online models, but setup, quality, and privacy differ by provider. | Partial support |
| Scheduled agents and automationsCustom agents and scheduled automations are documented platform features. | Supported |
Community signals
- Stars
- 37.5k
- Forks
- 2.5k
- Open issues
- 162
- Last commit
- 2mo ago
- Latest release
- 2.0.0-beta.28
- Repo created
- Aug 2021
Includes open pull requests
6mo ago
Refreshed nightly from the GitHub API.
Questions
What is Khoj used for?
Khoj is used as a personal AI assistant for searching documents, chatting with notes, researching the web, and running custom agents or automations.
Can you self-host Khoj?
Yes. The official docs provide a self-hosting setup path and position Khoj as a privacy-focused personal AI that can run on consumer hardware.
What files can Khoj search?
The official docs list PDF, plaintext, Markdown, Org-mode, and Notion content, with additional data-source and client integrations documented separately.
Does Khoj work with local LLMs?
Yes, Khoj can work with local models as well as online providers. The exact model and hardware requirements depend on the runtime and features you select.
Is Khoj an alternative to ChatGPT?
Khoj overlaps with ChatGPT for assistant-style chat, but its differentiator is searching and using your own documents plus self-hosting. It is better framed as a personal knowledge assistant than a universal hosted chatbot replacement.
Can Khoj search Notion and Obsidian content?
Khoj documents a Notion integration and an Obsidian client, while the import and synchronization behavior should be checked in the current integration guides.
Does self-hosted Khoj keep prompts private?
It can reduce exposure when local models and local data sources are used, but prompts or document context may leave the server if you configure an online model or external integration.
support // the lab
Found this write-up useful?
If it saved you time or a rebuild, you can support more practical homelab guides.