Oracle announced on July 14 that its AI Agent Studio for Fusion Cloud Applications now includes a developer-native builder experience, bringing Visual Studio Code, standard command-line tools, Git workflows, and AI coding assistants into the fold. The update allows customers and partners to build, test, and deploy multi-agent “Fusion Agentic Applications” directly inside Oracle Fusion — without leaving their familiar desktop development environments.

The move is a direct response to a familiar enterprise pain point: building AI prototypes outside the line-of-business platform is fast, but turning them into safe, governed production applications requires solving identity, data access, approvals, audit trails, and compliance controls from scratch. Oracle’s answer is to embed the entire AI runtime inside Fusion, inheriting its security and governance model.

What’s actually under the hood

The new capability is called AI Studio Skill. It’s a set of command-line interfaces, extensions, and workflows that lets developers use VS Code — the dominant editor on Windows and macOS — as the primary tool for creating agentic applications within the Oracle Fusion ecosystem. Instead of clicking through a browser-only low-code builder, a professional developer can now write, debug, and source-control agent logic using the same editor they use for every other project.

The toolkit supports:

  • VS Code extension for building and managing Fusion agents and multi-agent workflows.
  • Standard CLI for local validation, testing, and deployment.
  • Git-based lifecycle management, enabling team collaboration, branching, pull requests, and CI/CD pipelines.
  • AI coding assistants including OpenAI Codex and Claude Code, right inside the editor, to accelerate agent creation.
  • A soon-to-launch public GitHub repository with starter templates, sample applications, reusable assets, and reference architectures.

Alongside the pro-code path, Oracle continues to offer a no-code natural-language builder for business users. Both produce the same runtime artifacts — Fusion Agentic Applications — which are not simple chatbots but composite systems: specialized AI agents, user interfaces, workflow steps, approval gates, policy controls, and runtime assets tied together to deliver a specific business outcome, like accelerating financial close or reducing service escalations.

What it means for Windows developers and IT pros

For the Windows-based enterprise developer, the immediate gain is elimination of context switching. You can stay in VS Code on your Windows machine, use Git and your preferred CI/CD tools, and still target Oracle’s cloud platform. No need to log into a separate low-code environment for the AI parts. The agent project becomes just another artifact in your repository, testable locally before deployment.

For IT administrators and cloud architects, the bigger story is governance. Fusion Agentic Applications run natively inside Oracle Fusion Cloud Applications, meaning the runtime automatically enforces Fusion’s identity controls, data access policies, role-based approvals, and audit logging. When an agent updates a financial record or triggers a supply-chain action, it does so under the same permissions and visibility as a human user — and every step is auditable. This eliminates the risky “shadow AI” pattern where agents built outside the platform are later retrofitted with security controls.

Business power users who aren’t coders can still use the Agentic Applications Builder to create simple workflows via natural language, and those applications can later be refined by developers using the pro-code tools. It’s a unified framework with two doors, not two separate products.

Oracle says AI Agent Studio and the new builder experience are available at no additional cost to existing Fusion Applications customers and partners. The company already supplies over 1,000 pre-built AI agents and 22 Fusion Agentic Applications launched earlier in 2026. With the new tools, those can be extended or custom ones built from scratch.

The backdrop: from chatbots to governed multi-agent systems

Enterprise AI is undergoing a rapid shift from single-purpose copilots to multi-agent systems that can reason, coordinate, and execute. Microsoft’s own Copilot Stack and autonomous agent capabilities in Dynamics 365, Salesforce’s Agentforce, and SAP’s Joule are all chasing the same vision. What sets Oracle’s announcement apart is the emphasis on
native runtime governance and the deliberate inclusion of pro-code developer tooling.

Most platforms offer low-code agent builders, but they often fall short for enterprise development teams that need version control, automated testing, and repeatable deployments. Oracle’s previous AI Agent Studio was primarily a no-code environment. This update brings the same kind of tooling that developers expect from modern cloud platforms like Azure, AWS, or Google Cloud, but inside a SaaS ERP suite.

Another notable point: Oracle is not forcing developers into a proprietary language. You can work with standard code, use community coding assistants, and integrate with existing CI/CD pipelines. The public GitHub repo will further lower onboarding friction.

For Windows users, there is a subtle but important detail. VS Code on Windows is the default editor for countless IT professionals. The fact that Oracle specifically integrated with VS Code rather than a proprietary IDE signals that the company is serious about meeting developers where they are — on their Windows machines, with the tools they already trust.

What to do now if you’re a Fusion customer

If your organization already uses Oracle Fusion Cloud Applications, the new capabilities are available immediately at no extra cost. Here’s a practical checklist:

  • Check access: Log into Oracle Fusion and verify that AI Agent Studio is enabled in your environment. If not, an administrator can enable it from the setup console.
  • Install the extension: Search for the Oracle AI Agent Studio extension in the VS Code marketplace and install it on your Windows or macOS machine.
  • Set up CLI: Follow Oracle’s documentation to configure the command-line tools and authenticate to your Fusion instance.
  • Clone the starter repo: Once Oracle makes the public GitHub repository available, download a template that matches your use case — for example, a financial close agent or service escalation workflow — and use it as a sandbox.
  • Govern early: Before letting agents perform write operations (create, update, delete), work with your security team to review exactly which business objects the agents can touch and under what approval rules. Oracle’s documentation notes that agents can inherit Fusion’s access controls, but you must still configure those controls for the specific agent roles.
  • Test rigorously: Use local validation to simulate agent actions, then test in a non-production Fusion sandbox. The real challenge isn’t getting an agent to answer a question; it’s making sure the agent doesn’t wrongly modify a financial record or release a purchase order. Oracle’s testing framework inside AI Agent Studio can help build validation suites.

For partners and ISVs, the new pro-code path plus the marketplace expansion — from individual agents to full agentic applications — opens a packaging opportunity. You can now build, version, and publish reusable agentic applications on the Oracle AI Agent Marketplace, monetizing your domain expertise.

The Windows angle and what to watch next

This isn’t a Windows product update, but it directly impacts the large Windows developer community that works with enterprise cloud platforms. The integration of VS Code, Git, and standard CLI into a SaaS platform like Oracle Fusion could set a precedent. Expect other enterprise vendors — particularly those in ERP, HCM, and SCM — to offer similar pro-code extensions for their AI builders. For Windows power users and IT pros, the line between desktop development tools and cloud platform builders will continue to blur.

Oracle also hinted at a growing ecosystem: over 80,000 certified experts trained in AI Agent Studio are now available to help organizations build, test, and deploy. The combination of a certified talent pool, a VS Code–first development experience, and native governance could accelerate adoption of agentic applications significantly. Keep an eye on the public GitHub repository launch; it will be a bellwether for how quickly the community can produce real-world agentic solutions that go beyond demos.

For now, Windows-based enterprise developers have a new, pragmatic way to bring AI agents into production — not as standalone toys, but as fully governed parts of the core business system.