Prerequisites
- reference
Requirements — Python, a Couchbase cluster, uv or Docker, and an MCP client — you need in place before using the Couchbase MCP Server.
Before using the Couchbase MCP Server, make sure you have the following:
Required
-
Python 3.11—3.14: The Couchbase MCP Server aims to support Python versions in security or bug-fix (maintenance) status. Python versions that have reached their End-of-Life date are not supported.
-
A running Operational Insights cluster: A self-hosted Couchbase Operational Insights instance.
The Operational Insights MCP server supports only self-managed deployments. Capella-hosted Operational Insights is not supported. CompatibilityThe MCP Server is compatible with Operational Insights 2.0+. The following services are not supported: Sync Gateway, Couchbase Lite, and Capella AI Services.
-
uv or Docker: uv is the recommended way to run the server. Docker is an alternative if you prefer containerized deployments.
-
An MCP client: Such as Claude Desktop, Cursor, VS Code, Windsurf, or any MCP-compatible client.
Setup Operational Insights
If you do not already have an Operational Insights cluster, see the quick install guide to install one on your own infrastructure.
Getting Sample Data (Optional)
To try the server with sample data, import the travel-sample collections.
See Import the travel-sample collections.
Setup Authentication
Configure your Operational Insights cluster with one of the following authentication methods:
-
Basic Authentication: A username and password with access to the required databases.
-
mTLS Authentication: A client certificate and key for mutual TLS authentication.
For Basic Authentication setup, see Manage Users and Roles (self-managed).
For mTLS setup, see Configure Client Certificate Authentication.
Verify the following:
-
The cluster is accessible from the machine running the MCP server.
-
The database user has proper permissions to access at least one database.