Oracle Adds AI Studio Skill to Fusion 26C, Deploying Pro-Code Agents Inside Fusion Apps

Oracle is making AI Studio Skill available in Fusion release 26C, giving developers a pro-code route to create agentic applications that deploy directly inside Fusion Apps. The important distinction is not simply access to coding assistants: Oracle says the finished artifacts run in Fusion rather than as separately hosted AI applications that must be integrated and governed elsewhere.

A developer workflow aimed at the production environment

AI Studio Skill supports work through VS Code, Codex, Claude Code, Git, and the command line. Oracle supplies Fusion-specific knowledge, development patterns, and commands intended to create and manage Fusion Agentic Application artifacts from those familiar tools.

The target is a full application rather than an isolated agent. Oracle says a build can bring together specialist agents, workflows, tools, business objects, connectors, policies, approvals, enterprise context, communications, and the user experience that coordinates the work.

It is the pro-code counterpart to Agentic Application Builder, Oracle’s natural-language, no-code path released in Fusion 26B. Oracle says both paths use the same Fusion-native framework, separating the intended user rather than the environment in which the resulting application operates.

Testing, policy, and traceability are part of the build

Oracle is also packaging controls intended for business-process use alongside the code-generation workflow. The product description emphasizes checking an application’s behavior, setting the bounds for its actions, and retaining a record of what the system did.

• Testing and simulation can generate tests while an application is built, validate complete workflows, and use replay, synthesized data, and node-level inspection to examine behavior.
• Model optimization lets developers compare available models on quality, cost, and performance, then use lower-cost options where Oracle says results remain good.
• Deterministic workflows and policy nodes can encode business rules, standard operating procedures, and approval requirements; guardrails, human oversight, and traces cover model calls, tool use, data access, and business actions.

That combination addresses a practical difference between generating code and operating a system that can take business actions. Workflows and policy nodes provide a deterministic path for steps that must follow predefined rules, while tracing and approvals give teams mechanisms to inspect and constrain model-driven work.

The advantage is also a dependency

Running within Fusion gives these applications access to Fusion data, business objects, workflows, approvals, transaction services, identity, permissions, role-based security, observability, and auditability. It also means the pro-code path is deliberately tied to Fusion’s runtime and process infrastructure, rather than presented as a general-purpose route for deploying an agent anywhere.

Oracle has published a public GitHub repository, a README, a 26C setup guide, and instructions for incremental updates to support the workflow. Its announcement describes the architecture and controls, but does not provide customer deployment results or performance benchmarks, leaving the operational payoff to be established in use.