Khoj was founded in 2023 by Debanjum Singh Solanky and Saba Imran as an open-source personal AI assistant designed to serve as your “AI second brain.” The company, Khoj, Inc., is based in San Francisco, California and was backed by Y Combinator (W23 batch).
Khoj addresses a fundamental challenge in the AI era: how to leverage LLMs while keeping your data private and under your control. The founders recognized that:
- Data Privacy Concerns - Cloud AI services require sending sensitive data to third parties
- Vendor Lock-in - Proprietary platforms control your data, prompts, and conversation history
- Limited Customization - Hosted services offer limited flexibility for custom workflows
- Cost at Scale - Per-user pricing becomes expensive for teams and organizations
Khoj’s solution: a self-hostable platform that keeps data, prompts, and logs within your security boundary.
Khoj was built around Retrieval-Augmented Generation (RAG) from the start:
- Semantic search across your documents (PDF, Markdown, Notion, Word, org-mode)
- Context-aware responses using your knowledge base
- Web search integration for up-to-date information
Khoj emphasizes meeting users where they work:
- Web App - Universal browser access
- Obsidian Plugin - For note-takers using Obsidian
- Emacs Package - For Emacs users (khoj.el)
- Desktop App - Cross-platform native application
- Mobile - Phone-accessible interface
- WhatsApp - Chat via messaging app
From the beginning, Khoj supported multiple LLM providers:
- Cloud Models - OpenAI (GPT), Anthropic (Claude), Google (Gemini)
- Local Models - Ollama, llama-cpp, vLLM, LMStudio
- Open Models - llama3, qwen, gemma, mistral, deepseek
This flexibility ensures users aren’t locked into a single provider.
¶ Agent and Automation Features
Khoj evolved beyond simple Q&A to include:
- Custom Agents - Create agents with tunable personality, tools, and knowledge bases
- Scheduled Automations - Automate repetitive research tasks
- Deep Research Mode - Multi-hop reasoning and recursive web searches (
/research command)
- Code Execution - Terrarium sandbox for running Python code
- Image Generation - Text-to-image capabilities
- TTS & Voice - Text-to-speech and audio message support
¶ Growth and Milestones
| Milestone |
Details |
| Founded |
2023 (San Francisco, CA) |
| Seed Funding |
Y Combinator W23 batch |
| GitHub Launch |
Public repository opened |
| Current Stats |
33.3k stars, 2k forks, 5,148 commits |
| Releases |
170+ releases (latest: 2.0.0-beta.25, Feb 2026) |
| Community |
Active Discord, GitHub discussions |
Initial versions focused on:
- Basic chat interface with document retrieval
- Simple semantic search
- Single-model support
As adoption grew, Khoj added:
- Docker Compose Deployment - Simplified self-hosting
- Admin Panel - Django-based administration for model configuration
- Multiple Chat Models - Support for concurrent model configurations
- Notion Integration - Direct workspace sync
- Scheduled Tasks - Cron-like automation system
Modern Khoj includes:
- Agent Framework - Custom agent creation with tools
- Deep Research - Multi-hop reasoning agent
- Code Execution - Terrarium sandbox + E2B cloud option
- Computer Operator - Experimental computer control feature
- Multi-Modal - Image generation, TTS, voice messages
- Advanced RAG - pgvector-based semantic search
Khoj is released under the AGPL-3.0 license, which:
- Ensures modifications remain open-source
- Protects against proprietary forks
- Encourages community contributions
- Aligns with self-hosting ethos
This choice reflects the founders’ commitment to:
- Transparency - All code is auditable
- Community Trust - Users can verify security claims
- Sustainability - Community can maintain if needed
Khoj, Inc. operates as:
- Location: San Francisco, California
- Founders: Debanjum Singh Solanky (Co-Founder), Saba Imran (Co-Founder)
- Funding: Y Combinator W23 batch
- Business Model: Open-source core with potential cloud/hosted offerings
¶ Competitive Landscape
Khoj competes in the personal AI assistant and RAG platform space:
| Competitor |
Differentiation |
| OpenWebUI |
More focused on UI, less on agents/automations |
| AnythingLLM |
More enterprise-focused, different architecture |
| LocalAI |
Model serving focus, less on RAG/agents |
| Ollama |
Model runtime, not a full assistant platform |
Khoj’s differentiators:
- Full AGPL-3.0 open-source (not MIT/BSD)
- Deep research and agent capabilities
- Multi-platform clients (Obsidian, Emacs, Desktop)
- Scheduled automations
- Code execution sandbox
Khoj represents a privacy-first, self-hostable approach to personal AI. Its evolution from simple document Q&A to a full agent platform with deep research, automations, and multi-modal capabilities reflects the broader trajectory of the GenAI ecosystem.
Key principles that guide Khoj:
- User Control - Your data, your infrastructure, your rules
- Flexibility - Support for local and cloud models
- Transparency - Open-source codebase under strong copyleft license
- Accessibility - Meet users on their preferred platforms