HeadsUpAI

NVIDIA Launches AI-Q Agent Skill to Bring Deep Research to Coding Agents

NVIDIA packaged its AI-Q deep research pipeline into a portable agent skill. This allows developers to link the capability to harnesses like Claude Code or Codex, enabling these tools to offload complex research tasks to a local AI-Q server rather than processing them internally.
Pipeline stages
Intent classification, clarification, shallow research, and more
Supported harnesses
Claude Code, Codex, and OpenCode
Deployment options
Docker Compose and Helm charts
Data access protocol
Model Context Protocol
Authentication
OAuth2 and service accounts

While coding agents excel at orchestration, they often struggle with long-horizon research across massive datasets. This update follows the NVIDIA Agent Toolkit launch by turning a standalone blueprint into a modular component. Using the Model Context Protocol, AI-Q can now securely access authenticated enterprise data sources.

You can install the skill by linking the AI-Q repository to your local .claude/skills directory. The system uses a four-stage pipeline to produce structured reports with citations. The open-source blueprint is available on GitHub and supports deployment via Docker Compose or Helm charts for secure, on-premises environments.

View the full update on developer.nvidia.com
NVIDIA AI
NVIDIA AI
@NVIDIAAI
X

Say hello to open source deep research for your favorite agent harness. Our AI-Q agent skill packages the work of building a research pipeline into a portable skill. Drop it into your harness, and the agent delegates a research task to a local or hosted AI-Q server and gets back a detailed report with citations. See it in Codex below 👇

46retweets424likes
View on X

Still wondering? A few quick answers below.

The AI-Q Agent Skill is a portable version of NVIDIA's deep research pipeline designed to be integrated into agent harnesses like Claude Code and Codex. It allows general-purpose agents to delegate complex research tasks to a dedicated server, which then performs multi-source synthesis and returns a structured report with citations.

The pipeline uses a four-stage process to ensure research quality. It begins with an intent classifier to determine depth, followed by a human-in-the-loop clarifier to resolve ambiguity. A shallow researcher handles quick lookups, while a deep researcher performs long-horizon synthesis across enterprise data sources to produce a final cited report.

Yes, AI-Q is available as an open-source blueprint on GitHub. It is designed for flexible deployment where enterprise data lives, supporting Docker Compose and Helm charts. This allows the pipeline to run on developer laptops, on-premises Kubernetes clusters, or within air-gapped data centers to meet strict data sovereignty requirements.

AI-Q uses the Model Context Protocol to connect to enterprise data sources. It supports three primary integration patterns: unauthenticated servers, service-account authentication for batch jobs, and identity forwarding that preserves the signed-in user's bearer token. This ensures that research pipelines can pull from the same systems agents already use.

The AI-Q Agent Skill is currently validated for use with Claude Code, Codex, and OpenCode. Developers can install the skill by linking the AI-Q repository to their local skills directory. Once linked, the harness can route natural language research requests through a helper script that manages job submission and result retrieval.

Share this update