Model Context Protocol Server (MCP)

Since 26.0

This feature is available in Enterprise and AWS editions only.

CloudBeaver can run as a Model Context Protocol (MCP) server. This allows external AI clients to execute database operations through CloudBeaver.

Info

You can find more information about Model Context Protocol (MCP) here.

Enable MCP server¶

To activate the MCP server:

Navigate to Settings -> Administration -> Server Configuration
Enable the AI Integration and AI MCP Server options
Open the context menu for the connection where you want to use MCP and select Edit Connection.
Open the AI tab, select the Enable MCP checkbox, and click Save.

After you click Save, the MCP configuration section appears. It includes:
- the generated MCP Endpoint for the selected connection
- ready-to-copy mcp.json examples for supported MCP clients
  Example
```
{
  "mcpServers": {
    "dbeaver": {
      "type": "http",
      "url": "https://server-example.com/api/mcp/projects/General/datasources/postgres-jdbc-19b9378f1f8-60bb24cfcd2979a7",
      "headers": {
        "X-API-Key": "${token}"
      }
    }
  }
}
```
  Important
  The MCP endpoint is connection-specific. Each connection has its own endpoint and must be configured separately in the client.
- the Use OAuth authentication instead of API key option
  
  By default, snippets use API token authentication. To use OAuth, select Use OAuth authentication instead of API key. The generated snippets are updated automatically.
  
  Note
  OAuth support depends on the MCP client. If your client doesn’t support OAuth for MCP servers, use API token authentication instead.
If you use API token authentication, generate an API access token and replace ${token} with the generated token.

Tool	What it does
`getDatasourceInfo`	Returns DB type, version, and connection details.
`listCatalogNames`	Lists all catalogs/databases.
`listSchemaNames`	Lists schemas in a catalog.
`listSchemaNames`	Lists tables in a schema/catalog.
`getTableDetails`	Returns DDL for one or more tables (columns, constraints, indexes, comments).
`executeSQL`	Runs a SQL query and returns results as CSV (default `10` rows).

After you enable MCP for a connection, use the generated configuration snippet to register the server in your MCP client.

The snippet depends on the selected authentication method:

with API token authentication, the snippet includes the endpoint and token header
with OAuth authentication, the client handles authentication through the OAuth flow

Client configuration is performed entirely on the client side. Refer to your client’s documentation for instructions on adding an MCP server.

Client	Reference
Cursor	Cursor documentation
VS Code	VS Code documentation
Claude code	Claude code documentation
Claude Desktop	Claude Desktop documentation
ZED	ZED documentation

Tip

For other MCP clients, use the generated endpoint URL and adapt the configuration according to the client’s documentation.