By Siddartha Reddy Katukuri, HEXstream data integration engineer

Oracle Integration Cloud (OIC) is undergoing its most significant architectural evolution since launch. The shift from Packages to Projects is not just a UI reorganization, it is the gateway to AI Agents, RPA, advanced deployments, MCP server support, and every new capability Oracle is building into the platform.

All new OIC features will be delivered on Projects only. Staying on Packages means staying behind. So, flow the concrete, phased plan descibed below in order to migrate safely and future-proof your environment.

1.   Why Projects? The strategic case

Oracle built the concept of Projects around five pillars for next-generation automation, all orbiting a unified platform:

What a Project contains

Unlike a Package, which was simply an attribute of an integration, a Project is a first-class citizen with its own permissions, deployment pipeline, and artifact model. Here is everything a Project can hold:

Project vs. Package: capability comparison

2.   AI agents in OIC—a deep dive

AI agents are in OIC, but only if you're on Project. AI agents represent the leap from rule-based automation to goal-based automation. Within a Project, the AI agents section contains four building blocks:

•       Agents—the agent definitions that perceive, reason, decide and act

•       Tools—active integrations exposed as callable tools the agent can invoke

•       Agent patterns—thinking and reasoning strategies (built-in or custom)

•       Prompt templates—reusable prompt structures for consistent LLM interactions

How integration became an agentic tool

Any active integration within a Project can be promoted to an agentic tool. The agent then autonomously decides which tool to call and in what sequence should be used to achieve the designated business goal. With a single click, a Project can be exposed as an MCP (Model Context Protocol) server, giving external AI agents access to your enterprise integrations in a governed, auditable way.

LLM connectivity

OIC supports direct connections to multiple LLM providers, enabling flexible, best-of-breed AI strategies:

•       OCI Generative AI Service (Oracle-native)

•       Anthropic Claude

•       OpenAI (GPT-4 and variants)

•       Azure OpenAI

•       Custom API endpoints

Thinking patterns

Agents use configurable thinking patterns, reasoning strategies that determine how the agent breaks down goals, sequences tool calls, handles errors, and can decide when to escalate. Oracle provides built-in patterns, and teams can create custom ones to match specific business logic.

Observability for AI agents

Within the Project Observe tab, agent Instances are tracked in real time. Every action an agent takes is logged, making AI decisions explainable, governable, and recoverable—a critical requirement for enterprise deployments.

Oracle's design principle: every automated AI action must be explainable, governed and recoverable. This is built into OIC projects by default.

3.   Migration plan

The migration does not require a big-bang cutover. You can adopt at your own speed, one project at a time.

Recommended migration from Packages to Projects is structured across the sequence of steps below. Each action is designed so that the running integrations are unaffected throughout, and existing clients do not need to change anything until they are ready for them.

4.   Project-scoped observations

One of the most operationally significant advantages of Projects is scoped observability. Instead of filtering through all integrations in a global dashboard, each Project has its own observe tab with full visibility across every automation type it contains.

Instance-level insight

The instances tab within the observe section provides full drill-down: Primary identifier, Instance ID, Business identifiers, Status (Queued, In Progress, Succeeded, Errored, Aborted, Paused, Blocked, Resubmitted, Unknown), Received/succeeded timestamps, and Duration. A status overview donut chart gives an at-a-glance health signal.

Integration statistics are shown per version: Received, Processed, Succeeded, Errored, and Aborted counts, with a configurable time window and filtering by status and sort order.

5.   Key takeaways & action checklist

• Projects bring Co-Pilot, Agentic AI, RPA, Decisions, Deployments, and Repository support, none of which are available in Packages.
• Migration is safe and controlled. No big-bang required. Adopt at your own speed.
• Moving to Projects is future-proofing. All new Oracle Integration features will be Projects-only.
• The global forwarding pattern means existing clients are unaffected during migration.
• Project-scoped observability dramatically reduces noise in monitoring, which improves mean-time-to-resolution.

Your action checklist

READY TO MOVE FROM PACKAGES TO PROJECTS WITH A CLEAR, LOW-RISK STRATEGY? CLICK HERE TO CONNECT WITH US TO TURN YOUR ORACLE INTEGRATION TRANSFORMATION INTO A COMPETITIVE ADVANTAGE.