When speaking with IT leaders about their goals with agentic AI, the conversation has shifted from building an AI agent to managing and scaling all the AI agents they’ve built and inherited across many agent ecosystems. This is clearly a widespread change, as research shows that the number of specialized agents has more than doubled in the first half of 2025 alone, with projections suggesting over one billion active agents worldwide by 2029.
This explosion of agentic adoption has triggered a critical challenge for IT teams: agent sprawl.
In the rush to capture AI’s potential, departments have moved at different speeds using different environments. Your sales ops team might be deploying in Salesforce Agentforce, while engineers are building in Amazon Bedrock, and HR is piloting capabilities in Google Vertex AI.
When these autonomous tools are built in isolation, they become shadow agents – invisible to you, unaccountable to central security protocols, and disconnected from the broader enterprise strategy. You can’t govern what you can’t see, and for the 64% of IT leaders reporting being concerned about meeting their AI implementation goals, this lack of visibility is a primary roadblock to scaling.
To remedy this, you need a foundation of trust for your entire agentic strategy based around Agent Discovery. This automated process locates every AI agent in your IT estate, regardless of where it was built. At the heart of this process are Agent Scanners, which proactively scan your distributed cloud environments to:
- Eliminate shadow AI: Automatically locate every agent in your estate, regardless of the provider
- Normalize metadata: Extract and standardize the DNA of the AI agent. It identifies which Large Language Model (LLM) it uses, what specific skills it is authorized to perform, and which data servers it touches
- Establish a live registry: Create a continuous, real-time inventory that updates as new versions are deployed, ensuring that governance is never out of date
| Join Vijay at the whiteboard for a technical breakdown of why Agent Scanners are the critical foundation of any CIO’s agent discovery strategy. |
Organizations must implement a comprehensive AI agent management strategy to govern and orchestrate every agent across your business. MuleSoft Agent Fabric brings order to this growing complexity, providing the connective tissue for the agentic enterprise through a four-step framework: Discover, Orchestrate, Govern, and Observe.
- Discover with Agent Scanners: Use MuleSoft’s automated scanners to hit provider APIs, uncover hidden agents, and register them instantly into a centralized agent registry
- Orchestrate via Agent Broker: Once your agents are discovered, the next step is orchestration. You can connect disparate tools and agents into a unified Agent Broker, an intelligent routing service that dynamically organizes agents into business-focused domains
- Govern through AI Gateway for MCP and A2A Support: Next, these agents are no longer operating in a vacuum – they’re fully governed. Because they’re in the registry, you can apply policies to any cross-ecosystem agent. This ensures that they’re following your enterprise security and identity standards
- Observe with Agent Visualizer: Gain a real-time map of your agent network. Trace decision-making processes, monitor confidence scores, and detect hallucination risks to optimize your digital workforce and eliminate overlapping processes
The goal isn’t to slow down innovation, but to provide the guardrails that make innovation sustainable and scalable. By implementing a centralized identity and governance layer, you can allow departments to build on their platforms of choice while maintaining the security and visibility required to ensure your agents are working safely and securely for your employees.
To discover more about how to manage your AI ecosystem and maximize your ROI, explore MuleSoft Agent Scanners and MuleSoft Agent Fabric.
