Designing autonomous workflows

Individual agents are useful; coordinated agents are transformative. This lesson introduces how the MuleSoft Agent Fabric orchestrates many agents into reliable, autonomous workflows - the first of the capabilities you will practice in Track 3. (Recommended: complete Track 1 first.)

What you will learn

• What “guided determinism” means and why it matters for autonomy.
• The role of the Agent Broker in routing work between agents.
• How Agent Network 2.0 describes an agent network declaratively.

Guided determinism

Fully free-form agent behavior is powerful but unpredictable; rigid hard-coded flows are predictable but brittle. Guided determinism is the middle path: agents reason freely within boundaries you declare, so workflows stay flexible and governable (Get Started with Agent Fabric).

The Agent Broker

The Agent Broker is an intelligent routing agent that directs a request to the right downstream agent(s) based on their advertised capabilities, rather than you wiring every path by hand. It is the orchestration hub of the fabric.

Agent Network 2.0 (declarative)

You describe the network - its agents, their links, and the broker - as declarative YAML, so the topology is version-controlled and reviewable.

# agent-network.yaml (excerpt)
broker:
  id: enterprise-broker
  strategy: guided-determinism
agents:
  - id: triage-agent
    capabilities: [classify, route]
  - id: billing-agent
    capabilities: [invoice, refund]
links:
  - from: triage-agent
    to: billing-agent

Next steps

Continue to APIs as an AI control plane to see how governed APIs become the actions your orchestrated agents can safely take.

References

• Get Started with Agent Fabric