Guides

How the pieces fit together

The Marmot Development Kit handles authentication, account selection, and queries against your Komodo data in Snowflake. Web login and machine-to-machine credentials are both supported with automatic token refresh where applicable. The Model Context Protocol MCP server lets AI assistants explore your warehouse via tools such as list_snowflake.

You can also use App Builder to scaffold, build, and deploy containerized apps on the platform.

Guides

1. Quickstart — Authenticate, select an account, and run your first query (after install)
2. Authentication — Accounts, web login, M2M credentials, profiles
3. Executing Queries — Snowflake access, DB-API 2.0, CLI
4. Pandas Integration — read_sql, batches, analysis
5. Async Queries — execute_query_async and concurrency
6. MCP Server — Connect AI assistants to your data
7. App Builder — Build and deploy containerized apps
8. Troubleshooting — Common errors and fixes
9. Durable workflows (Restate) — Platform-managed Restate runtime for App Builder apps

Learning path

New to the SDK? Install, then follow guides 1 → 2 → 3; add 4 or 5 as needed.

Want AI-assisted exploration? After the quickstart, set up guide 6 (MCP Server).

Automating with CI/CD? Focus on guide 2 (Authentication) for service principals and profiles.

Building apps on the platform? Configure the MCP server, then follow guide 7 (App Builder); add guide 9 (Durable workflows) for crash-resilient workflows.

Something broken? See guide 8 (Troubleshooting).