Agentic workflows¶

Agentic workflows let you build long-lived AI services that run inside QPAQUE's confidential workspaces. Instead of writing orchestration code, you connect nodes—agents, retrievers, language models, redaction services, and more—on a canvas and launch them as a reusable, policy-governed, attested workflow. Once live, the workflow stays available and can be triggered via the Opaque SDK or API.

Each step in the workflow is represented as a node. Nodes can call models, retrieve data, apply guardrails, or transform text. Some nodes come preconfigured through integrations; others are configured directly by the workflow creator.

When to use agentic workflows¶

Use agentic workflows when you need AI to work with sensitive data in a repeatable way, not just as a one-off model call. They’re a good fit when you want to:

• Expose a persistent AI service that stays online and responds to structured requests.
• Control how data is accessed and transformed using retrieval, redaction, and guardrails.
• Prove what ran and under what conditions with attestation reports and audit logs.

Common examples include retrieval-augmented generation over private documents, governed conversational agents, and multi-step decision flows that combine several models or tools.

Getting started¶

To begin:

• Set up your local environment.
• Learn how to build, review, and launch a basic workflow.
• Explore how each node type works.
• Create or use shared model and data integrations.