Couchbase

Couchbase MCP Server is a self-hosted Model Context Protocol server that lets AI agents connect to and work with data in Couchbase clusters. It works with clusters hosted on Capella and with self-managed clusters.

The server exposes tools grouped into categories: Cluster Health, Data Schema, Key-Value, Query, and Performance. Cluster Health tools check server configuration, test the connection and report cluster health and services. Data Schema tools list buckets, scopes and collections and read the schema for a collection. Key-Value tools get, insert, upsert, replace and delete documents by ID. Query and Indexing tools run and explain SQL++ queries, list indexes and return index advisor recommendations. Performance tools report the longest running, most frequent and least selective queries, along with queries that use primary indexes or large responses.

Safety is built in. A read-only mode stops agents from changing data, and individual tools can be disabled so that an agent only sees the operations you allow. The server supports both STDIO and Streamable HTTP transports, so it can run locally beside an assistant or as a networked service.

The package is distributed on PyPI and as a Docker image on Docker Hub. The source is on GitHub under the Apache 2.0 licence. Setup steps, architecture notes, tool reference and configuration options are in the product documentation.

The server is aimed at developers and data teams who want AI assistants to read or change Couchbase data, or to diagnose slow queries and indexes, without giving the assistant unrestricted access.

Enterprise support is available by licensing Couchbase AI Data Plane. That licence also covers use and enterprise support of Couchbase Agent Memory and Couchbase Agent Catalog. The page does not list a price for the MCP server itself.