Oracle launches new AI Agent Marketplace for Fusion suite
Oracle has announced new updates to its AI Agent Studio for Fusion Applications, introducing an AI Agent Marketplace, enhanced support for leading large language models (LLMs), and expanded resources for AI development and deployment.
The new Oracle Fusion Applications AI Agent Marketplace provides customers with access to Oracle-validated, partner-built AI agents. These agents are integrated directly within Oracle Fusion Applications, allowing businesses to deploy automation capabilities and address industry-specific challenges without leaving their existing workflows.
Partner templates available in the marketplace are directly accessible through the Oracle AI Agent Studio. Customers can install, test, and manage both Oracle-certified partner templates and pre-built Oracle agents from a unified location.
The marketplace features agent templates developed by a range of system integrators and independent software vendors, enabling automation across finance, HR, supply chain, and customer experience. Some contributors from the Oracle PartnerNetwork include Apex IT, Grant Thornton, Huron, IBM Consulting, and Infosys. Additionally, Accenture, Deloitte, KPMG, and PwC have additional templates available to joint customers.
For example, IBM's Smart Sales Order Entry Assistant agent uses natural language prompts to streamline the sales process, reducing errors and manual steps during order creation and validation, thereby improving overall customer satisfaction. Additionally, KPMG's Purchase Order Item Price History agent delivers quick access to historical purchase data, helping procurement teams optimise negotiations by providing insights such as previous suppliers, purchase dates, and average prices.
Agent Management and Workflow Enhancements
The update introduces prompt libraries and lifecycle management capabilities, letting users store and manage all agent prompts and use cases centrally. A topics management function improves consistency by providing visibility into prompt boundaries for agents operating within similar domains.
Natalia Rachelson, Group Vice President, Cloud Applications Development at Oracle, told TechDay the applications are no-code offerings. The tools to build a customised agent from scratch are offered within the studio. Customers can modify additional agents and implement third-party relationships through MCA and A2A protocols.
Oracle's AI Agent Studio now features retrieval-augmented generation (RAG) support, allowing agents to incorporate information from various formats such as documents, images, and tables in their analyses. Agents can also perform RAG over external content repositories, including SharePoint, broadening their data access capabilities.
Further workflow improvements include the ability to set deterministic execution for business-critical processes, chain together multiple agent workflows for complex tasks, add agent nodes to workflows, and incorporate human-in-the-loop controls for oversight where needed.
Expanded LLM and Integration Support
The studio now supports a broader range of LLMs, including OpenAI, Anthropic, Cohere, Google, Meta, and xAI. This multi-model support gives customers flexibility in selecting the models best aligned to their business requirements.
Rachelson said that Oracle has a unique position in the market as it owns tech infrastructure that moves through the entire stack. She pointed to Oracle Cloud Infrastructure (OCI) as being purpose-built for AI and the cloud framework for these key LLM players.
"We have access to these LLMs so our engineering teams can tune them, can work with them, can ensure we don't have to get in line to get a token allotment. We have direct access to the LLMs through OCI Oracle Cloud infrastructure, and that allows us to go very fast," she said. "We haven't had to acquire companies to deliver AI and stitch it all together. It's all natively built within Fusion."
Other new technical capabilities include Model Context Protocol (MCP) integration for expanding agent connectivity with external systems, cross-agent collaboration through A2A agent cards, and a secure credential store that enables agents to manage authentication details for accessing APIs and third-party services.
The platform's observability has also been enhanced. Customers can now use a monitoring dashboard to review agent performance data in real time, conduct systematic performance evaluations, and capture detailed data for workflow analysis and troubleshooting. The addition of token usage monitoring helps customers manage costs associated with premium LLM usage.
