Privacy & Security

Privacy is a cornerstone principle of Khoj. Whether you use Khoj Cloud or self-host, we’re committed to keeping your personal data secure and under your control.

Privacy by Design

Khoj is built with privacy as a core feature, not an afterthought:

Always Self-Hostable

Run Khoj entirely on your own hardware with zero data leaving your network.

Open Source

All code is public and auditable on GitHub.

No Data Selling

We will never sell your data or use it to train models.

Minimal Collection

We collect only what’s necessary to provide the service.

Self-Hosted Privacy

Self-hosting gives you complete control and maximum privacy:

Complete Data Ownership

Your Hardware, Your Data

When you self-host Khoj, all your data stays on your machine:

Documents are stored in your PostgreSQL database
Embeddings generated locally never leave your device
Chat history remains on your server
No external servers see your queries or files

Offline Operation

Khoj can run completely offline:

docker-compose.yml

environment:
  # Use local models only
  - OPENAI_BASE_URL=http://localhost:11434/v1/
  - KHOJ_DEFAULT_CHAT_MODEL=qwen3

With local models like Ollama, you never need an internet connection.

No Telemetry Required

Disable anonymous usage telemetry completely:

docker-compose.yml

environment:
  - KHOJ_TELEMETRY_DISABLE=True

Telemetry helps us prioritize features, but it’s completely optional and anonymous.

What Gets Sent Where?

Even when self-hosting with online AI models, here’s what happens:

Data Type	Stored Locally	Sent to AI Provider	Sent to Khoj
Your documents	✅	❌	❌
Embeddings	✅	❌	❌
Chat messages	✅	✅ (Only relevant context)	❌
Usage stats	✅	❌	⚠️ (Anonymous, if enabled)

When using commercial AI models (OpenAI, Anthropic, Google), relevant portions of your indexed data may be sent as context with your queries. This is necessary for the AI to answer questions about your documents.

Secure Remote Access

Access your self-hosted Khoj securely:

Tailscale (Recommended)
Reverse Proxy
Local Network Only

Use Tailscale for encrypted, private network access:

End-to-end encrypted via WireGuard
No ports exposed to the internet
Access from any device on your Tailscale network

# Install Tailscale on your server
curl -fsSL https://tailscale.com/install.sh | sh

Use nginx or Caddy with HTTPS:

docker-compose.yml

environment:
  - KHOJ_DOMAIN=khoj.yourdomain.com
  - KHOJ_DJANGO_SECRET_KEY=your-secure-key
  - KHOJ_ADMIN_PASSWORD=strong-password

See Remote Access guide for details.

Access only from your home network:

docker-compose.yml

environment:
  - KHOJ_DOMAIN=192.168.1.100
  - KHOJ_NO_HTTPS=True

Ensure your router firewall blocks external access.

Khoj Cloud Privacy

Not everyone can self-host, so we’ve built Khoj Cloud with strong privacy protections:

Data Handling

What We Store

Your Documents & Embeddings

Stored in encrypted PostgreSQL database on AWS
Sharded by unique user ID (isolated per user)
Embeddings generated by open-source models on our private Hugging Face endpoints
Raw text stored to improve syncing and provide chat context

Your Account InformationWith Google SSO:

Name
Email address
Profile photo URL

We do NOT access:

Your Gmail
Google Drive
Any other Google services

Chat History

Stored encrypted in our database
Used only to provide conversation continuity
Never used for model training

What We Don't Store

IP addresses (anonymized in logs)
Detailed usage patterns (only aggregate metrics)
Your API keys for third-party services
Billing information (handled by Stripe)

Third-Party Services

When you use Khoj Cloud, your queries may be sent to:

Service	When	What’s Sent	Purpose
OpenAI	Using GPT models	Query + relevant document context	Generate AI responses
Anthropic	Using Claude models	Query + relevant document context	Generate AI responses
Google	Using Gemini models	Query + relevant document context	Generate AI responses
Hugging Face	Always	Document text only	Generate embeddings (on private endpoint)
Serper/Exa	Using `/online`	Search query only	Web search results

You control which AI models you use via Settings. Choose based on your privacy preferences.

Infrastructure Security

Encryption at Rest

All data in our PostgreSQL database is encrypted using AWS RDS encryption:

AES-256 encryption
Automatic encrypted backups
Encrypted storage volumes

Encryption in Transit

All communication uses TLS 1.3:

HTTPS for web traffic
Encrypted API calls
Secure WebSocket connections

Isolated Embeddings

Document embeddings are generated on stateless Hugging Face endpoints:

No persistent memory
Private dedicated endpoints
Hosted on AWS within our infrastructure

Access Controls

Multi-factor authentication available
Role-based access (for team plans)
Session management with secure cookies

Privacy Controls

You control your data:

Delete Your Data

Go to Settings → Danger Zone → Delete AccountImmediately removes all your data from our systems.

Export Your Data

Go to Settings → Data ExportDownload all your documents and chat history.

Revoke Integrations

Disconnect Notion, GitHub, or other integrations anytime.Stops data syncing immediately.

Choose AI Models

Select which AI providers you’re comfortable with.Different providers have different privacy policies.

Telemetry & Analytics

Khoj collects minimal, anonymous usage data to improve the product:

What We Collect

{
  "event": "chat_message_sent",
  "timestamp": "2024-03-05T10:30:00Z",
  "khoj_version": "1.0.0",
  "server_id": "anonymous-hash-xyz123"
}

What We DON’T Collect

❌ IP addresses
❌ Query contents
❌ Document contents or names
❌ Chat message contents
❌ File paths or folder structure
❌ Any personally identifiable information

How Telemetry is Used

Feature Prioritization: Understanding which features are most used
Performance Monitoring: Detecting performance issues across versions
Error Tracking: Identifying bugs to fix

All telemetry is sent to PostHog, an open-source analytics platform. View our telemetry code:

Disable Telemetry

Self-Hosted
Cloud

Add to your environment configuration:

docker-compose.yml

environment:
  - KHOJ_TELEMETRY_DISABLE=True

Or for pip installation:

export KHOJ_TELEMETRY_DISABLE=True
khoj

AI Model Privacy Comparison

Different AI providers have different privacy policies:

Provider	Data Retention	Training on Your Data	Privacy Policy
OpenAI	30 days	❌ (API zero retention)	Link
Anthropic	No retention	❌ (Explicit opt-out)	Link
Google Gemini	No retention	❌ (Enterprise API)	Link
Local (Ollama)	Never leaves device	❌	N/A

For maximum privacy, use local models with Ollama or LM Studio.

Security Best Practices

Follow these guidelines when using Khoj:

For Self-Hosting

Strong Admin Credentials

Set secure passwords:

docker-compose.yml

environment:
  - [email protected]
  - KHOJ_ADMIN_PASSWORD=use-a-long-random-password-here
  - KHOJ_DJANGO_SECRET_KEY=generate-a-unique-secret-key

Use a password manager to generate strong passwords.

Regular Updates

Keep Khoj updated:

# Docker
docker-compose pull && docker-compose up

# Pip
pip install --upgrade khoj

Network Security

Use Tailscale or VPN for remote access
Don’t expose ports directly to the internet
Use HTTPS with valid certificates
Keep firewall rules restrictive

Backup Your Data

Regularly backup your PostgreSQL database:

# Docker volume backup
docker-compose exec database pg_dump -U postgres postgres > khoj_backup.sql

For Cloud Users

Enable MFA

Add two-factor authentication to your account (coming soon).

Review Integrations

Periodically check Settings → Integrations and remove unused connections.

Monitor Activity

Check your recent conversations and uploaded documents regularly.

Compliance & Certifications

Khoj is working toward SOC 2 Type II compliance. Contact [email protected] for enterprise security requirements.

Current Status

Standard	Status	Notes
GDPR	✅ Compliant	Right to deletion, export, and access
SOC 2	🔄 In Progress	Expected Q3 2026
HIPAA	❌ Not compliant	Not recommended for healthcare PHI

Privacy Policy & Terms

Read our full legal documents:

Data Breaches & Incident Response

In the unlikely event of a security incident:

We’ll notify affected users within 72 hours
Provide details on what data was affected
Outline remediation steps taken
Publish a post-mortem on our blog

Your Data is Yours

Core Principle

We exist to serve you, not to monetize your data.Khoj is a sustainable, open-source alternative to closed-source corporate AI. We make money through subscriptions and enterprise licenses, not by selling user data.

FAQ

Can Khoj employees see my data?

On Khoj Cloud: Employees can only access your data for support purposes with your explicit permission or in response to legal requirements.Self-Hosted: Only you have access to your data.

Is my data used to train AI models?

No. We never use your documents, chats, or queries to train any AI models. This applies to both cloud and self-hosted deployments.

What happens if I delete my account?

All your data is permanently deleted within 30 days:

Documents and embeddings
Chat history
Account information
Integration credentials

Backups are purged after 30 days.

Can law enforcement access my data?

We comply with valid legal requests. For cloud users, we’ll notify you unless legally prohibited. Self-hosted users control their own data and legal obligations.

Is end-to-end encryption possible?

Not currently for cloud. E2E encryption would prevent server-side search and embeddings generation. Self-hosting provides equivalent privacy by keeping all data local.

Contact & Concerns

Privacy questions or concerns?

Email: [email protected]
Security Issues: [email protected] (PGP key available)
General: [email protected]

Interested in helping us build more privacy features? Join our Discord or contribute on GitHub.

Get Started

Features

Clients

Data Sources

Advanced

​Privacy by Design

Always Self-Hostable

Open Source

No Data Selling

Minimal Collection

​Self-Hosted Privacy

​Complete Data Ownership

​What Gets Sent Where?

​Secure Remote Access

​Khoj Cloud Privacy

​Data Handling

​Infrastructure Security

​Privacy Controls

Delete Your Data

Export Your Data

Revoke Integrations

Choose AI Models

​Telemetry & Analytics

​What We Collect

​What We DON’T Collect

​How Telemetry is Used

​Disable Telemetry

​AI Model Privacy Comparison

​Security Best Practices

​For Self-Hosting

​For Cloud Users

​Compliance & Certifications

​Current Status

​Privacy Policy & Terms

​Data Breaches & Incident Response

​Your Data is Yours

Core Principle

​FAQ

​Contact & Concerns

Build docs developers (and LLMs) love

Privacy by Design

Self-Hosted Privacy

Complete Data Ownership

What Gets Sent Where?

Secure Remote Access

Khoj Cloud Privacy

Data Handling

Infrastructure Security

Privacy Controls

Telemetry & Analytics

What We Collect

What We DON’T Collect

How Telemetry is Used

Disable Telemetry

AI Model Privacy Comparison

Security Best Practices

For Self-Hosting

For Cloud Users

Compliance & Certifications

Current Status

Privacy Policy & Terms

Data Breaches & Incident Response

Your Data is Yours

FAQ

Contact & Concerns