Skip to main content

​
Overview

Install Oxy, scaffold a project, and launch the web interface. You’ll need an OpenAI API key and a terminal (on Windows, use WSL 2).
1

Install Oxy

bash <(curl --proto '=https' --tlsv1.2 -LsSf https://get.oxy.tech)
You may need to restart your terminal for the oxy command to be recognized.
# Edge build (latest from main)
bash <(curl --proto '=https' --tlsv1.2 -LsSf https://nightly.oxy.tech)

# Specific stable version
OXY_VERSION="0.5.20" bash <(curl --proto '=https' --tlsv1.2 -LsSf https://get.oxy.tech)
Browse all releases at oxy-hq/oxy-nightly.
2

Create a project

Run in an empty directory:
oxy init
This generates a config.yml, a sample agent, workflow, data, and semantic layer file.
3

Set your API keys

The sample project is configured to use OPENAI_API_KEY and ANTHROPIC_API_KEY. Create a .env file following .env.example:
cp .env.example .env
Then fill in your API keys. For alternative models, see integrations.
4

Launch the web interface

oxy start
This starts PostgreSQL in Docker and launches the Oxy web UI.
Requires Docker. See the prerequisites guide if you don’t have a container runtime installed.
export OXY_DATABASE_URL="postgresql://user:pass@host:port/database"
oxy serve
Use oxy serve for production deployments or environments where Docker is not available.