Chain agents into automated workflows

FAQ

Frequently asked questions

What is multi-agent orchestration?

Multi-agent orchestration chains specialized AI agents into an end-to-end workflow — for example research → draft → review → publish. Each agent has its own role, model, and tools, and the next agent receives the previous agent's output. AgentWorks runs these as named pipelines you can trigger manually or on a schedule.

How do AgentWorks multi-agent pipelines work?

You compose a pipeline by adding agent steps in order. Each step picks its own LLM (GPT, Claude, Gemini, Mistral), its own knowledge base, and its own tools. Outputs flow into the next step. Pipelines run inside the same workspace as chats and tasks, so the result stream and audit log are unified.

Can I schedule multi-agent workflows?

Yes — on Pro and higher tiers. Schedule pipelines to run daily, weekly, monthly, or at a custom time. Each scheduled run debits the workspace wallet with full per-step token transparency.

What's the difference between an AI agent and a multi-agent pipeline?

A single AI agent is one model + one set of tools doing one task (e.g. "draft a sales email"). A multi-agent pipeline is multiple agents stitched into a single workflow, where each step refines or extends the previous one — closer to how a human team operates.

Are multi-agent pipelines EU AI Act compliant?

AgentWorks is EU AI Act-ready, not blanket "compliant" — the risk class depends on your use case. Every step in the pipeline inherits the same EU AI Act controls: per-step risk classification, an optional PII-detection layer at the gateway, a complete audit log of every model call, and human-in-the-loop approval where required.