Marmot is an open-source data catalog and context layer for AI, developed by Marmotdata. It helps organizations connect to and catalog various data assets across their data stack, including databases, tables, topics, and data pipelines. Marmot provides unified data discovery, ownership queries, and business glossary lookups via its built-in MCP server, RESTful API, and CLI. It supports AI clients like Claude and ChatGPT, making it suitable for data teams, individual contributors, engineers, and AI agents seeking powerful data discovery.

What does Marmot connect to?

Marmot connects to and catalogs diverse data assets across an organization's data stack. This includes various databases, individual tables, topics, message queues, and APIs. It also integrates with cloud storage solutions, dashboards, and data pipelines. This broad connectivity allows Marmot to serve as a central context layer for AI. Its operation requires PostgreSQL as the sole external dependency, which it uses for search capabilities, job scheduling, and storing all metadata.

Core features and tools

Marmot exposes key callable tools through its built-in Model Context Protocol (MCP) server. These include discover_data for unified data discovery, find_ownership for bidirectional ownership queries, and lookup_term for business glossary lookups. Beyond the MCP server, Marmot also offers a RESTful API for programmatic interaction and a command-line interface (CLI) for broader management. Both the MCP server and REST API utilize streamable HTTP as their transport protocol.

Who is it for?

Marmot serves data teams and individual contributors who need powerful data discovery without complex enterprise infrastructure. It is also designed for engineers and AI agents. It aids users in finding data assets, tracing data flows, and managing metadata. This platform supports cataloging data assets, enriching them with business context, and maintaining shared glossaries. Skip this if your organization cannot or will not deploy PostgreSQL as an external dependency.

How to get started

The server authenticates using an API key passed as an environment variable. The MCP server and REST API support streamable HTTP as the transport protocol.

Use cases

• Data Discovery and Understanding: Find any data asset quickly using full-text search, structured queries, and metadata filters, answering questions like "What tables do we have related to customer orders?".
• Impact Analysis and Lineage Tracing: Trace data flows and analyze the impact of changes, with AI agents capable of querying ownership to understand dependencies, such as "What breaks if we rename the order_gmv column?".
• Metadata Management and Collaboration: Catalog data assets, enrich them with business context, assign ownership, and maintain shared glossaries, with assets populated via CLI, API, or Terraform.

Marmot offers a data catalog and context layer for AI, connecting many data assets and providing tools for discovery, ownership queries, and glossary lookups. It caters to data teams, engineers, and AI agents, supporting crucial workflows like data discovery and impact analysis. Authentication is flexible, supporting API keys, OAuth, and SSO. Start by generating an API key from your user profile and querying for customer order tables.