# MotherDuck

> Serverless cloud data warehouse built on DuckDB with isolated per-user and per-agent compute, MCP access, and agent-built dashboards and pipelines.

MotherDuck is a serverless cloud data warehouse powered by DuckDB, aimed at data teams, app developers, and AI agents. It gives each user, dashboard, or agent session its own isolated DuckDB instance, which MotherDuck calls a Duckling. Users can query through SQL, Python, a CLI, a REST API, a Postgres endpoint, or MCP-compatible agents.

## What It Is

MotherDuck is a managed, cloud-hosted version of DuckDB for analytics. It runs SQL analytics in a serverless warehouse, and the product pages describe dual execution that can run query stages on a laptop, in the cloud, or across both. Databases can be shared through read-only, zero-copy shares, and MotherDuck can query data in S3 and other S3-compatible object storage. Storage options listed include Parquet, Apache Iceberg, DuckLake, differential storage, and cloud object storage.

## Hypertenancy and Isolated Compute

Hypertenancy is MotherDuck's term for giving each user, customer, or AI agent an isolated DuckDB instance. According to the company, this avoids shared clusters and noisy neighbors, and cold starts land under 100ms. Instances can be sized from Pulse through Giga, and the product page mentions up to 16 read-scaling replicas. User-level CPU visibility and cost attribution are described as built in.

## Agent-Native Tooling

The MCP Server lets agents such as Claude, ChatGPT, and Cursor inspect catalogs, run SQL, read Guides, create Dives, and manage Flights, with answers returned as readable SQL in sandboxed compute. Three product components support this workflow:

- **Dives:** interactive visualizations built by an agent or with SQL, which can be shared or embedded in an application.
- **Flights:** scheduled Python pipelines that run on MotherDuck compute for ingestion, enrichment, and transformation.
- **Guides:** shared business context that agents use to produce more accurate queries.

Agent Skills and a CLI are also offered.

## Integrations and Ecosystem

MotherDuck lists integrations across ingestion (Fivetran, Airbyte, dlt), BI (Tableau, Power BI, Looker, Hex, Omni, Metabase), transformation (dbt), and orchestration (Airflow, Dagster, Prefect, Kestra). Language clients include Python, Node.js/TypeScript, Go, Java, Rust, R, C++, WebAssembly, and ODBC.

## Features
- Serverless DuckDB cloud data warehouse
- Hypertenancy: isolated DuckDB instance (Duckling) per user, customer, or agent
- Dual execution across local and cloud
- Zero-copy database shares
- MCP Server for AI agents
- Dives: agent-built interactive visualizations
- Flights: scheduled Python pipelines
- Guides: shared business context for agents
- DuckLake and Apache Iceberg support
- Postgres endpoint
- Read-scaling replicas
- User-level CPU visibility and cost attribution

## Integrations
dbt, Hex, Tableau, Power BI, Looker, Omni, Metabase, Fivetran, Airbyte, dltHub, Apache Airflow, Dagster, Prefect, Kestra, Claude, ChatGPT, Cursor, Amazon S3, PostgreSQL, Jupyter

## Platforms
WEB, API, CLI, DEVELOPER_SDK

## Pricing
Freemium — Free tier available with paid upgrades

## Links
- Website: https://motherduck.com
- Documentation: https://motherduck.com/docs/getting-started/
- EveryDev.ai: https://www.everydev.ai/tools/motherduck