|Documentation
dataworkers

Deployment

Data Workers can be deployed as a local development tool, a Docker container, or a Kubernetes service. This page covers each deployment option.

Local Development (Default)

The simplest deployment is running the MCP server locally via npx:

npx dw-claw

This requires Node.js 20+ and npm. The server runs as a local process and connects to your MCP client (Claude Code, Cursor, etc.) via stdio transport. No Docker or infrastructure required.

Docker

Run Data Workers as a Docker container for isolated, reproducible deployments:

docker pull ghcr.io/dhanushashetty/data-workers:latest
docker run -it --env-file .env ghcr.io/dhanushashetty/data-workers:latest

Pass infrastructure credentials via environment variables using --env-file or individual -e flags:

docker run -it -e SNOWFLAKE_ACCOUNT=xxx -e SNOWFLAKE_USER=xxx -e SNOWFLAKE_PASSWORD=xxx ghcr.io/dhanushashetty/data-workers:latest

Docker Compose — For multi-service setups, create a docker-compose.yml:

services:\n data-workers:\n image: ghcr.io/dhanushashetty/data-workers:latest\n env_file: .env\n stdin_open: true\n tty: true

Run with docker compose up.

Kubernetes (Planned)

Kubernetes deployment via Helm charts is planned for a future release. The intended deployment model:

  • •Helm chart with configurable agent selection
  • •Resource requests and limits per agent
  • •ConfigMap-based environment variable management
  • •Horizontal scaling for high-throughput environments
  • •Health check endpoints for liveness and readiness probes

Join our Discord at discord.com/invite/b8DR5J53 or watch the GitHub repo for announcements when Helm charts are available.

Environment Variables

All deployment methods use the same environment variables. Set these to connect to real infrastructure:

  • •SNOWFLAKE_ACCOUNT, SNOWFLAKE_USER, SNOWFLAKE_PASSWORD — Snowflake
  • •GOOGLE_CLOUD_PROJECT, GOOGLE_APPLICATION_CREDENTIALS — BigQuery / Dataplex
  • •DATABRICKS_HOST, DATABRICKS_TOKEN — Databricks
  • •DBT_API_TOKEN, DBT_ACCOUNT_ID — dbt Cloud
  • •AIRFLOW_HOST, AIRFLOW_USER, AIRFLOW_PASSWORD — Apache Airflow
  • •DAGSTER_HOST, DAGSTER_TOKEN — Dagster Cloud
  • •PREFECT_API_KEY, PREFECT_API_URL — Prefect Cloud
  • •DW_LICENSE_KEY — License key for Pro/Enterprise tier (enables write tools)

Without credentials, the server uses InMemory stubs with sample data. No external services required to get started.

For MCP client configuration (Claude Code, Cursor, etc.), see the MCP Client Setup page and the dedicated platform guides in the Setup section.

ConfigurationFAQ