Skip to main content

AI agents are only as good as the context they can reach.

Marmot is the open source context layer. It catalogs what you run, what it means and who owns it, then serves that to your team in the UI and to every assistant over one MCP server.

Get startedLive demo

Works with

Claude
Cursor
Copilot
Gemini
ChatGPT
Windsurf

One layer between your systems and everyone asking about them

Plugins keep it filled. People and agents read from the same place.

PostgreSQL
Kafka
S3
dbt
Trino
Iceberg
MarmotMarmotContext layer
AssetsLineageOwnershipGlossary
MCP server
Web UI
REST API
CLI
Slack

Your rows never move. Marmot holds what exists, what it means and who owns it, and you can also push that in from Terraform, Pulumi or the API.

Answers, not guesses

Questions that used to land in a team's Slack channel get answered on the spot, from your catalog instead of a schema someone pasted into a prompt. One MCP server, not one per source.

Set up MCP

What tables do we have related to customer orders?

discover_data

I found 3 assets matching "customer orders": the orders table in the warehouse, an orders_raw Kafka topic, and a daily_orders_summary view.

What breaks if we rename the order_gmv column?

get_lineage

The daily_orders_summary view and the Revenue Overview dashboard both depend on that column. The table is owned by the Data Platform team, so check with Sarah Chen first.

Not just for agents

The catalog your assistants read is the one your team searches, browses and edits.

Two services. Not seven.

Marmot is one binary and a Postgres database. Other catalogs want Elasticsearch, Kafka, Neo4j, MySQL and Airflow running before they show you a single table.

Quick start
~ / marmot
$docker compose up -d
[+] Container postgres-1 Started
[+] Container marmot-1 Started
Marmot is running at http://localhost:8080

Our largest open source deployment serves 175+ active users. Load tested well beyond that:

500k+Assets
100+Concurrent users
<50msAvg response time
How we load tested it

Plugins for what you already run

More than 30 of them, from databases to queues to warehouses.

PostgreSQLMySQLMongoDBClickHouseBigQueryKafkaAirflowdbtS3TrinoDuckDBElasticsearchKubernetesSQSRedis

View all plugins or ask for one that isn't here yet.

Metadata only

Schemas, ownership and lineage. Your rows, messages and payloads never enter Marmot.

Self-host or Cloud

Run it inside your own VPC under your own controls, or let us run it for you.

MIT licensed

Built in the open. Read exactly what it collects and what it stores.

One context for people and agents

Try the live demo, or deploy it yourself in a few minutes.

Live demoTalk to us

Built by engineers from HashiCorp, Adidas, Just Eat Takeaway.com and Traefik, who help maintain Kubernetes, Terraform, Redpanda and the Cloud Native Computing Foundation.