As organizations begin deploying artificial intelligence (AI) agents, IT teams must help them move beyond traditional custom applications toward becoming an agentic enterprise. This shift represents more than just another technology upgrade. It’s a fundamental reimagining of how businesses operate, orchestrate workflows, and deliver value.
While the opportunities are exciting, organizations need to be aware of and seek to prevent agent sprawl. Similar to API sprawl, where IT teams struggle to manage the proliferation of APIs, the concept is quickly becoming an issue for AI agents. IT teams need a seamless way to manage — discover, orchestrate, govern, and observe — all the agents within the enterprise.
“We need to think of an agent as a new species within the ecosystem,” explains Prashant Choudhary, field CTO for MuleSoft. “Agents can communicate using natural language and they have the ability to plan and reason, which means enterprises can accomplish more things faster because they no longer must deal with the rigidity of custom apps.”
That’s where an agent fabric comes into play. It’s a sophisticated ecosystem, building on a composable infrastructure, where AI agents collaborate seamlessly across business domains. It uses natural language to orchestrate complex workflows — enabling coordination, cooperation, and soon-to-be competition among agents.
This transformation turns existing APIs, workflows, and events into intelligent tooling that agents can dynamically access and combine to fulfill business requests. To make this transformational vision a reality, IT needs to implement an agent fabric.
The architecture of agent collaboration
An agent fabric provides the foundation for the agentic enterprise. There are four key components to this cohesive, intelligent fabric:
- Discover: Create a catalogue or registry that acts as a directory service for the entire agent ecosystem and is available to all agents. It includes a list of all agents, as well as their domains, skills, and capabilities
- Orchestrate: Establish an agent broker that enables agents to collaborate effectively
- Govern: Enable policy enforcement
- Observe: Allow IT to understand complex agent actions and interactions across the entire environment
Each business unit — finance, customer service, human resources, procurement, etc. — has its own headless domain agent with department- and function-specific knowledge. These agents communicate through the Agent2Agent (A2A) protocol, which enables them to discover each other, understand their respective capabilities, and coordinate across organizational boundaries.
An agent broker facilitates collaboration between agents, enabling loose coupling rather than rigid, predetermined workflows. For example, when a fulfillment agent needs assistance, it can dynamically identify and engage with order management, inventory, and shipping agents as needed, creating flexible, context-aware workflows.
Governance and observability at scale
As organizations deploy hundreds or thousands of agents across the enterprise, IT management becomes increasingly complex.
A well-designed agentic strategy addresses this challenge. MuleSoft, for example, enables organizations to develop and embed governance policies that intersect every message entering and leaving the system, enabling real-time policy enforcement without manual intervention. This is crucial given that agents operate probabilistically, requiring more sophisticated security and compliance controls than traditional deterministic applications.
Also, observability becomes essential when managing complex agent interactions that may span 15 or more system layers. MuleSoft offers advanced visualization tools that provide workflow transparency, so IT teams can understand how agents interact and accomplish tasks through the entire system.
The economic impact
By enabling agents to understand intent and automatically orchestrate the appropriate resources, organizations can respond more quickly to market changes — and even create entirely new markets — by reducing operational overhead and freeing human talent for higher-value strategic work.
The agent fabric provides an infrastructure layer that makes this transformation possible, creating an ecosystem where agents from different vendors can collaborate. This interoperability ensures that organizations aren’t locked into single-vendor solutions while maximizing the value of their existing technology investments.
MuleSoft Agent Fabric turns uncontrolled agents into a governed and highly coordinated, intelligent network, enabling you to govern any agent anywhere, orchestrate agents across ecosystems, and observe every agent interaction.