SnowCrew AI

SnowCrew AI is an open-source, visual no-code studio for building, running, and exporting multi-agent CrewAI workflows. Design agents, tasks, and crews on a drag-and-drop canvas, wire them to Snowflake Cortex LLMs and data plus integrations like GitHub, then export ready-to-run crew.py / YAML. Run it locally for development, or deploy it to Snowflake Container Services (SPCS) for production.

What It Does

• Creates local workspaces for CrewAI planning and export
• Models agents, tasks, crews, and Snowflake connections
• Supports Snowflake API-style config
• Exports starter agents.yaml, tasks.yaml, crew.py, .env.example
• Stores app data locally in ~/.crewai-studio-local/ (overridable via DATA_DIR for container/SPCS deployments)

Running It

Development (Local)

For local development and trying it out:

git clone https://github.com/Blake-Daniel-Inv/SnowCrew-AI.git
cd SnowCrew-AI
npm install
npm run dev

Then open http://localhost:3000.

To verify a production build locally:

npm run build
npm start

Production (Snowflake Container Services)

In production, SnowCrew AI runs as a Snowpark Container Services (SPCS) service inside your Snowflake account. The high-level flow:

1. Provision the Cortex role, database, image repository, stage, compute pool, and secrets. Run snowflake/setup-cortex-role.sql and snowflake/setup.sql — see docs/OPERATIONS.md §1 for the full walkthrough, including the SNOWFLAKE_PAT and CREDENTIALS_MASTER_KEY secrets (snowflake/PAT-SETUP.md).
2. Authenticate to your account's image registry:

   docker login <org>-<account>.registry.snowflakecomputing.com -u <snowflake_user>

3. Build, push, and deploy the service:

   ./snowflake/deploy.sh <image_repository_url>

4. Wait until the service is ready, then health-check it:

   SELECT SYSTEM$GET_SERVICE_STATUS('CREWAI_STUDIO_SVC');  -- wait for READY

   curl -sf "https://<your-ingress-host>/api/health/ready"

The service spec is snowflake/spec.yaml, and app data is persisted to DATA_DIR (a mounted stage) rather than ~/.crewai-studio-local/. The complete runbook — secrets, GitHub OAuth setup, backups, secret rotation, and debugging — lives in docs/OPERATIONS.md.

Integrations

SnowCrewAI agents can call out to third-party APIs on a per-user basis. The first integration is GitHub (read-only access to private repos, pull requests, and issues).

To enable it:

1. Register a GitHub OAuth App at https://github.com/settings/developers. Set the callback URL to https://<your-host>/api/auth/github/callback (or http://localhost:3000/api/auth/github/callback for local dev).
2. Set the following environment variables before starting the server:
   • GITHUB_OAUTH_CLIENT_ID — the App's Client ID
   • GITHUB_OAUTH_CLIENT_SECRET — the App's Client Secret
   • CREDENTIALS_MASTER_KEY — a 32-byte key (base64 or hex) used to envelope-encrypt stored tokens
3. Restart npm run dev (or your production server).
4. Open /settings and click Connect GitHub. After authorizing, your agents that declare the github tool will pick up your token automatically the next time their crew runs.
5. After connecting, your GitHub organizations are listed so you can see which org repos your agents have access to.

License

Released under the MIT License.

Acknowledgments

Built on top of CrewAI, the open-source multi-agent framework (MIT licensed). CrewAI is installed separately in your Python environment and is not redistributed as part of this project.