Skip to main content
A project is the unit of work inside an Oxygen workspace — it holds your agents, procedures, data apps, and the config.yml that wires them to data sources and models.
1

Create the project

From app.oxygen-hq.com, open your workspace and click + New project in the sidebar. Give it a name and Oxygen scaffolds a starter config.yml for you.
2

Connect a model

Open Manage → Models (or edit config.yml directly in the IDE) and add the LLM you want to use. The secret comes from Manage → Secrets.
config.yml
models:
  - name: openai-4.1
    vendor: openai
    model_ref: gpt-4.1
    key_var: OPENAI_API_KEY
3

Connect a data source

Add the warehouse the agents will query. The simplest option is to drop a CSV or Parquet file into the project and let Oxygen’s in-cloud DuckDB engine read it; switching to BigQuery, Snowflake, Postgres, or any other warehouse is a one-block change.
config.yml
databases:
  - name: local
    type: duckdb
    dataset: ./
See Connect your data for the full decision tree across warehouses.
4

Save and proceed

Save the config.yml. Oxygen validates the file and surfaces any schema errors inline.
Now you’re ready to build your first agent, or let the Builder copilot do it for you in Build with chat first.