Dev.to AI πŸ€– Ai πŸ‘ 0 πŸ“– 2 min read

Microsoft and Oracle connect Copilot agents to Oracle databases

Microsoft and Oracle announced a set of integrations at Microsoft Ignite that connect AI agents to data held in Oracle databases. The details come from a post on the Azure Blog, co-authored with two Oracle executives and

Microsoft and Oracle announced a set of integrations at Microsoft Ignite that connect AI agents to data held in Oracle databases. The details come from a post on the Azure Blog, co-authored with two Oracle executives and published on November 17.

What is new

The integrations are built on Oracle AI Database@Azure. They let agents reason over Oracle data from finance, supply chain, HR and customer operations, together with the context Microsoft Work IQ draws from meetings, email, documents and Teams. Users reach this through Excel, Teams and Microsoft Copilot.

Three pieces are concrete:

  • Excel plug-in. The Oracle AI Database plug-in for Excel is available now. Users connect a workbook directly to Oracle Database, pull data into the sheet and analyze it with Copilot. The post says this removes the export step and the IT ticket.
  • Azure Workload Center for Oracle. This is in limited preview. It groups a workload's architecture, resources and dependencies into a single object called a Workload Virtual Instance. Together with Azure Monitor Health Models and the Azure Copilot Run Oracle agent, it links monitoring signals from the application, database and infrastructure layers.
  • Managed MCP servers for developers. Agents built in Microsoft Foundry can connect to Oracle AI Database through managed Model Context Protocol servers and the A2A protocol. Access is limited to approved SQL and PL/SQL operations, tools and predefined reports, whether the database runs in Oracle Cloud Infrastructure or as Oracle AI Database@Azure.

How access is controlled

Oracle AI Database@Azure supports Microsoft Entra authentication and Oracle Deep Data Security. According to the post, that keeps agents inside each user's existing permissions. Application roles decide which tools a user or group can use, and agent actions remain auditable across the Oracle estate. The post also says the data stays in place: the agent goes to the database, and the data is not copied out to the AI.

Scenarios Microsoft describes

The post gives three examples. An admin investigates a slow order-processing application by prompting Copilot, which coordinates several agents, each limited by that admin's identity, and returns one operational conclusion. An analyst has Copilot examine Oracle transaction data, add context from Work IQ, flag exceptions, draft an explanation and prepare an action plan for review. A developer builds a sales intelligence app in which separate agents pull account and order data, suggest cross-sell options and assess account health, then deliver a short sales brief through an API. A customer service agent grounded through Foundry IQ can also answer policy or account questions with citations and recall history from Oracle-backed memory.

Other details

Oracle AI Database@Azure now covers 33 Azure regions. Oracle has said FedRAMP Class D certification for US government and public-sector workloads is coming soon. Proximity Placement Groups, which place a database nearer to its applications, are described as cutting VM-to-Exadata network round-trip latency by about 50% compared with deployments without that alignment.

The source does not state pricing, licensing terms or a general-availability date for the Workload Center preview.

πŸ“° Read the original article on Dev.to AI

Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes β€” full credit and traffic to the original publisher.