NVIDIA publishes guide to deploying AI-Q Blueprint on Oracle Cloud Infrastructure

NVIDIA has published instructions for deploying AI-Q Blueprint, an open platform for long-running AI agents, in a production environment on Oracle Cloud Infrastructure.

From question-and-answer to the long-running agent

In two years, agentic AI has gone through three generations. The first systems worked in a "one question, one answer" mode: the model did not remember previous steps. Then came multi-turn dialogue, in which the model began to retain context within a session, allowing coherent conversations and iterative refinement of results. Today's long-running agents work on a fundamentally different principle. They can:

• set goals and break them into subtasks
• delegate steps to specialized sub-agents
• keep full context over the course of a long task
• run external tools, such as a browser, a code interpreter and APIs, in a secure sandbox
• adapt the plan on the fly if intermediate results diverge from expectations

What NVIDIA AI-Q Blueprint is

AI-Q Blueprint is an open reference design that NVIDIA is developing for teams building agentic systems. The framework implements a long-running agent architecture out of the box and provides ready-made patterns for orchestration, tool management and long-term context storage.

The project is fully open: companies can adapt the Blueprint to industry-specific tasks, from software development automation to data analytics and customer support. Instead of building an agent architecture from scratch, a team takes a proven foundation and builds its own logic on top of it.

Deployment on Oracle Cloud

Oracle Cloud Infrastructure supplies the Blueprint with GPU capacity and managed services: storage, networking, task isolation and monitoring. Together they form a stack in which the agent gets the resources it needs at the right moment, without manual cluster management.

The guide's keyword is production-ready. This is not a demo: the configurations are designed for real workloads, with automatic scaling and enterprise-grade security. For teams moving from a prototype to a working system, the Blueprint cuts deployment time from several months to several weeks.

Why this changes the market

Just a year ago, launching a long-running agent required deep MLOps expertise: custom task queues, state management, tool isolation. Only large technology teams could handle that work. AI-Q Blueprint combined with Oracle's cloud infrastructure changes the equation: an average enterprise team gets a ready-made architecture that it only needs to configure for its own data and business processes. That lowers the barrier to entry and speeds the shift of agentic AI from "lab experiment" to "working product."

What this means

Long-running AI agents have matured enough to leave research labs. The pairing of NVIDIA's open framework with Oracle's cloud platform cements a new standard: GPU stack + managed cloud + open Blueprint is now the standard route to production agents for companies without an in-house AI infrastructure team.