Prerequisites for MCP Server

To use MCP-compatible AI clients with Autonomous AI Database through MCP endpoints, a developer must ensure the following: the correct database version, privileges, authentication credentials, network and security controls, and tool registration readiness.

Supported Database and Version

MCP Server is supported on Autonomous AI Database 26ai and 19c. See About MCP Server for more details.

Supported MCP Clients

You must use AI agent applications that support remote MCP over streamable-http transport and OAuth 2.1 or token-based authentication. This enables your client applications to interact with MCP tools. MCP Server uses streamable HTTP with SSE as the transport channel. Client applications can connect to MCP server by using database credentials or bearer token (MCP servers supports OAuth or OIDC and token based authentication) so that they can see the available tools. See Configure MCP Server in AI Agent Application for more details.

See:

Privileges to Enable MCP Server

You must have the OCI permission to update the Autonomous AI Database to add the required free-form tag for enabling MCP server.

Authentication and Authorization

Use your database credentials for authentication so that AI clients can identify you and securely access the MCP server. Autonomous AI Database MCP server supports MCP clients that use OAuth 2.1 and token based authentication.

For MCP clients supporting bearer token, a bearer token for their database can be generated using an API. You can then add it in the MCP server configuration. See Configure MCP Server in AI Agent Application for more details.

Custom Tools

You must have custom tools that are registered correctly so that MCP clients can list and call them with the appropriate roles and least-privilege access. Only privileged database users can register tools. You can create custom tools using the DBMS_CLOUD_AI_AGENT.CREATE_TOOL procedure. See Example: Create Select AI Agent Tools and CREATE_TOOL Procedure for more details.