TLDR: Oracle has adopted the Model Context Protocol (MCP), an open standard, to enable AI agents to securely access and interact directly with data stored in its flagship database. This integration, implemented via Oracle SQLcl, simplifies agentic AI development by allowing AI assistants to not only generate SQL queries but also execute them and evaluate results, fostering more autonomous and efficient AI workflows. The move positions Oracle as a key player in AI-driven database management, emphasizing security and standardization.
Oracle has announced its adoption of the Model Context Protocol (MCP), a significant step that will allow artificial intelligence (AI) agents to securely and directly access data within its widely used database systems. This integration, which was implemented around mid-July 2025, marks a pivotal moment for AI-driven database management and agentic AI development.
The Model Context Protocol, an open standard introduced by generative AI vendor Anthropic, is designed to standardize how AI agents connect with external resources such as databases and large language models (LLMs). Prior to MCP, developers often had to write custom code for AI models to communicate with various data sources, a process that was complex and time-consuming. MCP streamlines this connectivity, acting as a “universal connector” that enables AI agents to interact with multiple systems uniformly.
Oracle is integrating MCP support into its command-line tool, SQLcl. This allows any AI assistant that supports MCP to establish direct database connectivity. The new functionality extends beyond merely generating SQL code; AI assistants can now execute queries directly and evaluate the results. This capability opens up new possibilities for “agentic workflows,” where AI agents can implement advice, assess consequences, and perform tasks autonomously. For instance, a developer can ask an AI assistant to explain a database structure, and the AI can generate, execute, and analyze SQL queries to provide the explanation.
Security remains a paramount concern for Oracle. The company emphasizes that AI access to databases requires the same rigorous security considerations as traditional access. The implementation via Oracle SQLcl ensures that existing security measures remain intact, with credentials managed locally and connections utilizing the same protocols as regular database interactions. For organizations already using SQLcl, MCP support requires no changes to their current security architecture. Oracle recommends using databases with minimal user privileges for AI assistants, avoiding production environments, and instead working with cleaned-up, read-only replicas to mitigate data exposure risks. All interactions are logged in a DBTOOLS$MCP_LOG table within the user schema, enabling monitoring and detection of suspicious activity.
Major tech companies like AWS, Google, Databricks, and Snowflake have already adopted MCP, highlighting its growing industry acceptance. By integrating MCP, Oracle positions itself as one of the first major database vendors to offer direct support for the protocol, potentially giving it a significant advantage in the evolving landscape of AI-driven database management. Analysts, such as Holger Mueller from Constellation Research, underscore MCP’s value in simplifying the training of agentic AI applications and standardizing their interoperability.
Also Read:
- Teradata Unveils Open-Source MCP Server to Enhance AI Agent Data Access
- ThoughtSpot Unveils Agentic MCP Server to Enhance AI Interoperability in Analytics
This move is part of Oracle’s broader strategy to embed AI capabilities directly into its database, as seen with Oracle Database 23ai, which includes features like AI Vector Search and Select AI for natural language queries. The goal is to make AI more accessible, scalable, and powerful by bringing it directly to where the data resides, eliminating the complexity of data movement and enabling non-technical users to interact with databases using natural language.
