Oracle adds MCP support to advance agentic AI development
Given that the open standard simplifies the complex process of connecting systems during agentic AI development, the tech giant's integration with MCP stands to benefit users.
As agentic AI development gains popularity, Oracle is the latest data management provider to add support for Model Context Protocol.
Created by generative AI (GenAI) vendor Anthropic, Model Context Protocol (MCP) is an open standard that lays out an accepted way AI agents connect with the resources that train them, such as databases and large language models (LLMs) and work in conjunction with other agents to act autonomously. The standard aims to simplify developing agents as well as ensure that they act safely and responsibly.
Tech giants such as AWS and Google now support MCP, as do data platform vendors Databricks and Snowflake, and data management specialists such as SnapLogic and StarTree, among many other vendors.
By enabling Oracle Databases to integrate with MCP servers, Oracle on July 16 became the latest to add support for the protocol. Because MCP simplifies training agentic AI applications and standardizes how they interoperate, MCP support is a valuable addition for Oracle Database users, according to Holger Mueller, an analyst at Constellation Research.
It is the widely adopted standard to tell AI models what can be accessed outside of what is stored, added and learned by them in training, which is super important for [training processes such as] retrieval-augmented generation.
Holger Mueller Analyst, Constellation Research
"It is the widely adopted standard to tell AI models what can be accessed outside of what is stored, added and learned by them in training, which is super important for [training processes such as] retrieval-augmented generation," he said.
Based in Austin, Texas, Oracle is a tech giant that competes with AWS, Google Cloud, Microsoft and others. Its broad array of data management and analytics capabilities includes Oracle Database, among numerous other database offerings.
Enterprises have increased their investments in building AI applications since OpenAI's November 2022 launch of ChatGPT. The chatbot was a significant improvement in generative AI technology, given GenAI's potential to make workers better informed and more efficient.
Initially, development was largely focused on chatbots that enable users to ask questions and receive responses in natural language. By mid-2024, agentic AI began to emerge.
Unlike chatbots that aid workers only when prompted, agents possess context awareness and reasoning capabilities that enable them to act autonomously. Among their capabilities are performing tasks such as searching data for insights and automating certain repetitive processes that take significant time when done manually.
However, because less human involvement is involved in agentic processes than in those of past AI tools, it is critical that agents are trained to act and interact in ways that don't put enterprises at risk of hurting themselves or their customers.
To standardize that training, Anthropic introduced MCP in November 2024. Subsequently, Google Cloud released Agent2Agent Protocol in April to similarly provide a standard for agentic interactions.
Without such protocols, the communication between systems that enable developers to build applications that combine LLM capabilities with proprietary data is complex and time-consuming to execute, according to William McKnight, president of McKnight Consulting Group.
"Without MCP, interactions have a language barrier, requiring a human interpreter to facilitate communication," he said. "MCP [enables] LLMs to directly understand and interact with databases [to make] the process more efficient, autonomous and scalable."
Given the importance of MCP, McKnight -- like Mueller -- noted the importance of its addition for Oracle users.
"This integration enhances developer productivity, simplifies integration efforts and lowers barriers for business users to access data insights using natural language interfaces," he said.
Oracle users can integrate MCP through the command-line interface (CLI) for Oracle Database. Once integrated, the CLI can be run as an MCP server and enable Oracle Database to securely connect their databases to MCP tools so they can train agents with the proprietary data that enables agents to understand their organization.
To address data exposure concerns that many organizations have when using their proprietary data with external systems, Oracle recommends that customers take caution when granting LLMs access to their data by replicating data or using dedicated data subsets. In addition, it advises users to regularly audit LLM queries to detect anomalies and attempts to access sensitive information and other restricted data.
Oracle Database's support for MCP is seemingly simple to set up and secure, according to McKnight.
"Oracle's implementation seems technically sound and designed for ease of deployment, leveraging existing tools like [the CLI] to enable direct interactions between LLMs and Oracle Database," he said.
However, whether it proves to be technically sound in practice remains to be seen, according to Mueller.
"It is early days," he said. "We will see how AI can get to the data. It is all version one [of integrations] for all vendors."
Oracle's motivation for adding MCP support, meanwhile, came from the potential value the open standard can provide enterprises that are trying to develop agentic tools, according to Jeff Smith, a distinguished product manager at Oracle.
"The industry has adopted MCP extremely quickly, and the potential was obvious from the beginning," he said. "We evaluated it internally and saw quite clearly that the potential for Oracle Database users was real. The key is doing it in a way that makes sense, is secure and will scale."
Looking ahead
While Smith declined to address Oracle's product development plans for Oracle Database, Mueller suggested that the most important thing the tech giant should do is continue adding capabilities such as MCP support that keep up with current trends.
Developments such as adding vector search and storage in May 2024 made Oracle Database one of the most versatile and powerful data management tools, according to Mueller. In fact, Oracle Database is so advanced compared with its competition that now AWS is partnering with Oracle to advance its own database capabilities, he noted.
Continuing to add capabilities will keep Oracle Database competitive.
"Keeping it modern and relevant [should be Oracle's focus," Mueller said. McKnight, meanwhile, suggested that Oracle expand its AI infrastructure to attract more customers interested in AI development. In addition, he suggested adding more to its MCP integration and said addressing the ease of use of its database tools could be beneficial.
"Oracle could focus on improving accessibility and ease of adoption for business users, making it simpler for non-technical individuals to query the database using AI interfaces," McKnight said.
Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.
