Deploying Cognee

Cognee is a standalone tool that can be deployed easily on-premise or in the cloud. This guide covers the various deployment options available and helps you choose the right approach for your use case.

Deployment Options Overview

Local Development & Testing

MCP Integration - Connect to code editors and IDEs
Docker - Containerized local deployment

Cloud & Production Deployments

Modal - Serverless deployment for easily scalable remote processing
Helm - Production-ready Kubernetes deployment
EC2 - Traditional cloud server deployment

When to Use Each Option

Option	Best For	Complexity	Scalability
MCP	Code editors, development workflows	Low	Limited
Docker	Local development, testing, small deployments	Low	Medium
Modal	Serverless processing, auto-scaling workloads	Medium	High
Helm/K8s	Production environments, enterprise deployments	High	High
EC2	Traditional server deployments, custom configurations	Medium	Medium

Quick Start

Choose your deployment path:

For Development

Start with MCP integration for code editor workflows:


docker run --env-file ./.env -p 8000:8000 --rm -it cognee/cognee-mcp:main
 
# Connect to your preferred editor
# See MCP integration guides for specific editors

For Production

For most production use cases, we recommend starting with Modal for its simplicity and automatic scaling, then moving to Kubernetes with Helm for enterprise deployments that require more control.

Docker Deployment

Docker provides an easy way to run cognee in a containerized environment, perfect for local development and testing.

Prerequisites

Docker installed and running
Docker Compose (for multi-service setup)

Environment Configuration

Configure your deployment using environment variables:


# Database connections
POSTGRES_URL=postgresql://user:password@localhost:5432/cognee
NEO4J_URL=bolt://localhost:7687
 
# Vector database
QDRANT_URL=http://localhost:6333
 
# LLM configuration
OPENAI_API_KEY=your_openai_key

Architecture Considerations

Single Instance Deployment

Docker or EC2 for simple, single-server deployments
Suitable for development, testing, or small-scale production use
Limited scalability but easier to manage

Distributed Deployment

Modal for serverless, auto-scaling processing
Kubernetes/Helm for container orchestration
Better for high-volume, production workloads

Database Requirements

All deployment options require:

PostgreSQL - Primary data storage
Vector Database - Qdrant, Weaviate, or Pinecone
Graph Database - Neo4j (optional but recommended)

Next Steps

Choose your deployment method based on your use case
Follow the specific deployment guide for detailed instructions
Configure your databases according to your requirements
Test your deployment with sample data

Need Help?

Join our Discord community for deployment support
Check our Troubleshooting Guide for common issues
Review the API Reference for configuration options

Ready to deploy? Choose your preferred method above and follow the detailed guides for step-by-step instructions.