marmotdata.io

What It Is

Marmot positions itself as an “Open Source Context Layer for AI.” In practice, it is a lightweight open-source data catalog. It helps teams register data assets such as database tables, Kafka Topics, Queues, APIs, and Pipelines; enrich them with ownership, business definitions, tags, custom fields, and a Glossary; and expose that context to engineers and AI tools through a UI, REST API, CLI, and MCP.

Core Capabilities

Functionally, Marmot covers data discovery, full-text and structured search, interactive lineage, upstream/downstream dependency analysis, data product grouping, and team collaboration. Its standout feature is the built-in MCP Server, which allows MCP-compatible clients such as Claude Desktop, Claude Code, Cursor, Cline, and LibreChat to query assets, lineage, ownership, and business terms in natural language. Architecturally, Marmot is built in Go, runs as a single binary, and depends only on PostgreSQL for search, task scheduling, and metadata storage. Compared with traditional data catalogs that require multiple components such as Elasticsearch and Neo4j, it is easier to deploy.

Integrations, APIs, and Self-Hosting

Marmot supports deployment via Docker Compose, Docker, Kubernetes/Helm, and CLI, with a Quick Start that demonstrates local setup in just a few minutes. Its ecosystem includes plugins for PostgreSQL, MySQL, MongoDB, ClickHouse, BigQuery, Kafka, Airflow, dbt, S3, DuckDB, SQS, OpenAPI, and more. It also provides a UI configuration wizard, CLI, REST API, Terraform, and Pulumi support. The documentation gives fairly detailed guidance on plugin permissions, configuration options, and collectible metadata. However, the current version is 0.9, and some plugins are marked Experimental, so stability should be validated before production adoption.

Pricing

The self-hosted edition is MIT-licensed, free, has no feature gates or usage limits, and includes no telemetry. It includes all plugins, MCP, REST API, CLI, search, lineage, and governance capabilities. Marmot Cloud is still under development and is planned to offer hosting, upgrades, backups, and a free tier. Professional services are priced based on project scope.

Pros, Cons, and Who It’s For

Its strengths are open-source transparency, simple deployment, and a clear approach to integrating AI context. It is well suited for data platform teams, data engineering teams, and organizations that want to connect a real data catalog to AI Agents. Limitations include the fact that the cloud version is not yet officially available, information on enterprise-grade governance, auditing, and commercial support is limited, and plugin maturity needs to be assessed one by one. The source text does not provide details on accessibility from China; dependencies such as GitHub, container images, and Discord may experience network instability in mainland China. If access is constrained, alternatives such as OpenMetadata, DataHub, and Amundsen may be worth evaluating.