Oracle Autonomous AI Database MCP Server is a multi-tenant, built-in feature of Autonomous AI Database Serverless that exposes Model Context Protocol (MCP) endpoints. It enables AI agents and clients – such as Claude Desktop and OCI AI Agent – to invoke tools you define with Select AI Agent. We are excited to announce the general availability of Oracle Autonomous AI Database MCP Server for Autonomous AI Database supporting database versions 19c and 26ai.

As Autonomous AI Database natively integrates with Oracle Database Identity, OCI IAM, and external OIDC/OAuth 2.1 identity providers for authentication and authorization, Autonomous AI Database MCP Server enables organizations to manage AI agents and client applications to discover permitted schemas, run SQL, and use database features—all designed to be enforced by your database’s security policies. The MCP server is fully managed, requiring no MCP setup, and delivers enterprise-grade auditing and performance controls for AI-to-data workflows.

Why MCP, and why now?

MCP standardizes how AI systems connect to external tools and data. Instead of custom adapters, MCP offers an open, universal, open interface for discovering tools, reading resources, and invoking actions. Oracle Autonomous AI Database MCP Server, now available in-database, helps organizations to manage AI agents that are designed to securely and reliably interact with Autonomous AI Database using a widely adopted, vendor-neutral standard.

What Makes Autonomous AI Database MCP Server Unique

Autonomous AI Database MCP Server is fully embedded within your database instances as a managed, Oracle-supported service. There is no need to deploy or maintain separate MCP infrastructure or servers—this can help reduce operational overhead.

Unlike third-party database MCP servers, the Autonomous AI Database MCP Server offers deep integration with native database security features. It is designed to respect and enforce your existing roles, lockdown profiles, access control lists, auditing, encryption, role-based access control (RBAC), and virtual private database (VPD) policies, aiming to deliver a level of protection that external MCP implementations cannot match.

Additionally, tools created using Select AI Agent are instantly accessible through the MCP Server, including those leveraging Select AI for natural language to SQL (NL2SQL) generation and retrieval augmented generation (RAG). For example, to process a natural language query, MCP servers typically require steps to access and enumerate database schemas, identify available objects, and gather metadata before the LLM can generate and run the SQL query. In contrast, NL2SQL Select AI-enabled tools within the Autonomous AI Database MCP Server can bypass this discovery process using an AI profile, allowing you to use different LLMs—potentially optimized for SQL quality—for streamlined, efficient NL2SQL query generation and execution.

For enterprise needs, denial-of-service protections and rate limiting are governed by your ECPU allocation. The multitenant architecture is built to centralize governance while help preserving policy, data, and performance isolation per tenant.

Autonomous AI Database MCP Server brings together Oracle’s advanced AI, security, and database technologies into a single, cohesive solution with the following benefits:

Operation, Architecture, and Integration

  • Native and Unified Deployment: Natively run MCP Server inside your Oracle Autonomous AI Database instance (with database versions 19c/26ai), with no need for standalone servers or cluster management, and directly manage it within the OCI console.
  • Seamless Experience and Support: Benefit from Oracle AI Database’s fully automated updates and patching with fully Oracle-managed services, including access to Premier Support and SLAs. Connect MCP-compatible clients (e.g., Claude and OCI AI Agent) and use Select AI Agent-defined tools.

Compliance, Security, and Performance

  • Enterprise-Grade Security and Compliance: Address compliance leveraging Oracle’s built-in auditing, data residency, and industry standards. Be able to achieve fine-grained, least-privilege access via roles, ACLs, lockdown profiles, and VPD.
  • Optimized In-Database Performance: Run AI tasks and queries in-database to minimize latency and data movement, with integrated resource management for predictable throughput.

Cost Optimization and Monitoring

  • Elastic Scaling and Automated Controls: Take advantage of Autonomous AI Database’s auto-scales for resiliency and high throughput, with automated throughput control (ECPU allocation) for cost and performance efficiency.
  • Comprehensive Monitoring: Access real-time usage metrics and auditing provided through OCI Monitoring and enterprise tools, with support for streaming incremental responses (SSE).

Getting started

Getting started with the Autonomous AI Database MCP Server is straightforward—just follow these four steps:

  1. Enable the MCP server
  2. Create Select AI Agent tools
  3. Choose and configure your MCP client / AI agent application
  4. Restart your AI application

1. Enable the MCP server

You can enable the MCP server by running the following OCI free-form tags as the ADMIN user on your Oracle Autonomous AI Database instance. Check your company’s AI and data security policies to confirm compliance. Enabling the MCP Server creates the MCP endpoint bound to the database OCID. After enabling the MCP server, the database exposes its MCP endpoint, which authenticated MCP clients can use to run registered tools:

First, sign in as the ADMIN user on your Oracle Autonomous AI Database instance, set the OCI free-form tag adb$feature with the following tag value:

Tag Name:  adb$feature
Tag Value: {"name":"mcp_server","enable":true}

Enabling the MCP Server creates the MCP endpoint that is bound to the database OCID:

http://dataaccess.adb.<region-id>.oraclecloudapps.com/adb/mcp/v1/databases/{database-ocid}

2. Create Select AI Agent tools

Create custom tools using the DBMS_CLOUD_AI_AGENT.CREATE_TOOL PL/SQL API. For example, you can create a tool to run a SQL query. In the example below, we create a custom tool named “MY_RUN_SQL_TOOL” and register it as a callable MCP tool. (For more examples, like list schemas and objects, or get object details, see the documentation.)

BEGIN
  DBMS_CLOUD_AI_AGENT.CREATE_TOOL (
    tool_name  => 'MY_RUN_SQL_TOOL',
    attributes => '{"instruction": "This tool runs the provided read-only (SELECT) SQL query.",
       "function": "RUN_SQL",
       "tool_inputs": [{"name":"QUERY","description" : "SELECT SQL statement without trailing semicolon."},
 	               {"name":"OFFSET","description": "Pagination parameter. Use this to set the page size when performing paginated data retrieval."},
                       {"name":"LIMIT","description" : "Pagination parameter. Use this to specify which page to fetch by skipping records before applying the limit."}
                      ]}'
        );
END;

Define the corresponding PL/SQL function, run_sql. The function accepts a SELECT statement and returns paginated JSON results. This supports large result sets while being designed to respect database security policies such as VPD.

CREATE OR REPLACE FUNCTION run_sql(
    query      IN CLOB,
    offset   IN NUMBER,
    limit    IN NUMBER
) RETURN CLOB
AS
    v_sql      CLOB;
    v_json     CLOB;
BEGIN
    v_sql := 'SELECT NVL(JSON_ARRAYAGG(JSON_OBJECT(*) RETURNING CLOB), ''[]'') AS json_output ' ||
        'FROM ( ' ||
        '  SELECT * FROM ( ' || query || ' ) sub_q ' ||
        '  OFFSET :off ROWS FETCH NEXT :lim ROWS ONLY ' ||
        ')';
    EXECUTE IMMEDIATE v_sql
        INTO v_json
        USING offset, limit;
    RETURN v_json;
END;

3. Choose and configure your MCP client / AI agent application

Because Autonomous AI Database MCP Server supports standard MCP, any MCP client can interface with the MCP Server, including VS Code with Cline, Claude Desktop, OCI AI Agent, custom applications, and others. Configure your MCP client with the MCP server endpoint created in step 1.

For example, the following snippet shows a sample configuration for Claude Desktop. After downloading and installing Claude Desktop from claude.ai/download, navigate to file à Settings à Developer and click “Edit Config”. This takes you to the claude_desktop_config.json file where you add a configuration similar to the following:

{
  "mcpServers": {
      "autonomous-database-mcp-server": { <- Customer-provided MCP server name
      "command": "/opt/homebrew/bin/npx", <- Executable/command invokes MCP server
      "args": [             <- Arguments to the command to connect to MCP server
        "-y",
         "mcp-remote",                        
        "http://dataaccess.adb.../databases/OCID1...OC1...7H", <- MCP Server URL
        "--allow-http"
      ],
      "transport": "streamable-http"
    }
  }
}

4. Restart your AI application

Restart the MCP client application—for example, terminate the Claude Desktop application process and reopen it. Then sign in with your database username and credentials. The application will display only the Select AI Agent tools that the authenticated user can access. Tools are selected automatically based on your natural language prompts.

Give it a try

Ready to get started? Take the Oracle Autonomous AI Database MCP Server for a test drive and experience how it can simplify integration, help enhance security, and accelerate AI-driven access to your data.

With standardized MCP tooling, enterprise‑grade controls, and built‑in observability, the MCP Server helps you confidently build and scale agent‑based applications on Autonomous AI Database—eliminating unnecessary glue code and complexity.

Resources

For more information…