Pune Meetup - MuleSoft Agent Fabric Inside

Pune MuleSoft Meetup
22-August-2026
Agent Fabric Inside: Agents, Broker, and
Governance

Safe Harbour Statement
● Both the speaker and the host are organizing this meet-up in individual capacity only. We
are not representing our companies here.
● This presentation is strictly for learning purposes only. Organizer/Presenter do not hold
any responsibility that same solution will work for your business requirements.
● This presentation is not meant for any promotional activities.
2
Pune MuleSoft Meetup Group

Speaker
Venkatesh is a MuleSoft mentor and working as a Lead
Consultant at Infosys. He has great experience in
Integration, development, execution and delivery using the
MuleSoft products and offerings. Venkatesh has completed
the following certifications –
• Integration Associate
• MCD Level 1 and 2
• MCPA and MCIA
• Hyperautomation Specialist
• Salesforce AI Associate
• Salesforce AI Specialist
✔ LinkedIn – https://www.linkedin.com/in/venkatesh-jujarao-6a4349106/
✔ Medium - https://medium.com/@vicksjujarao

Agent Fabric Inside: Agents, Broker, and
Governance
Introduction & Context
➢ Why MuleSoft Agent Fabric and When to Use It
Defining the Solution
➢ What is MuleSoft Agent Fabric
➢ Agent Network Framework
Business Requirement & Use Case
➢ Zero-Touch IT Operations
Technical Deep Dive
➢ Network Architecture
➢ Agent Network Project Structure & Components — Quick Look
Core Building Blocks of Agent Fabric (MCP, Bridge, A2A, ACB)
Prerequisites
Demo: Agent Fabric in Action
➢ Sequence of flows and steps

Why MuleSoft Agent Fabric and When to Use It ?
Organizations accumulate multiple MCP servers and agents.
Result: fragmented, siloed, hard to govern.
Agent & MCP Sprawl
Agents and MCPs often run without enterprise-grade security.
Risk: uncontrolled access, compliance gaps.
No Security & Governance
AI agents or systems deployed without IT awareness or approval.
Hidden workflows introduce unmonitored risks and compliance blind
spots.
Shadow AI
No clear audit trail of agent actions.
Hard to answer: “Who did what, when, and why?”
Lack of Traceability
Each agent acts standalone.
No orchestration, duplication of effort, inconsistent outcomes.
No Coordination

What is MuleSoft Agent Fabric ?
Agent Fabric helps you turn agent sprawl into a governed, coordinated, intelligent network — providing the infrastructure to
discover, orchestrate, govern, and observe agents across your enterprise.
🔍
Discover
C E N T R A L R E G I S T R Y
A central registry with automated
discovery where every agent and MCP
server — regardless of where it's built
— can be cataloged, discovered, and
reused.
Eliminates duplication · Compose from
existing assets
🛡️
Govern
E N T E R P R I S E G U A R D R A I L S
Enterprise-grade guardrails ensuring
every agent, MCP server, and LLM
interaction is secure, compliant, and
cost-controlled so teams can innovate
with confidence.
Security · Compliance · Cost Control
⚙️
Orchestrate
G U I D E D D E T E R M I N I S M
Coordination of agents and tools
across ecosystems with guided
determinism, so multi-step processes
run smoothly and every agent delivers
reliable outcomes.
Multi-vendor · Cross-platform · Reliable
📊
Observe
E N D - T O - E N D V I S I B I L I T Y
End-to-end visibility into how agents
make decisions and perform, turning
black-box AI into transparent,
accountable systems.
Decisions · Performance · Accountability
Together, these four capabilities transform disconnected agents and MCP servers into a unified, trusted, high-performing digital workforce.
Build and deploy Salesforce-native agents
for use within the Salesforce ecosystem.
Agentforce
The agent control plane for managing
and governing agents across any
platform
Agent Fabric
Agentic Enterprise
Architecture
Registry (Exchange)
Scanners
Agent Broker
Agent Network
Agent Visualizer
Monitor
Omni Gateway
Governance
Strategies
Cost Management

Agent Network Framework
Anypoint Code Builder
(ACB)
• Script
• Canvas
DEV
API Manager
• Governance
• Policy
• LLM Proxies
Security
Exchange
• Catalog
• Discovery
Repository
Runtime Manager
• Broker (Agent network
application)
Deployment
Omni Getaway
• Ingress
• Egress
Traffic
Observability
• Agent Visualizer
• Anypoint Monitoring
Operations
Agent Network
Framework

The Requirement: Zero-Touch IT Operations
Use case
WHAT WE NEED TO ACHIEVE
Zero-Touch IT Operations (AMS)
⚡ Auto-detect incidents — no manual triage
📋 Execute runbooks — without human steps
🗂️ Update CMDB — real-time, auto-synced
🔔 Notify & close tickets — end-to-end
WHAT WE HAVE
ServiceNow
MCP Server
Incidents · APIs · Runbooks
MuleSoft
MCP Server
APIs · Flows · Integrations
?
The missing piece
Agent Fabric →

Network Architecture
Notification Listener
App
Ingress
Omni Gateway
Egress
Omni Getaway
Agent Broker
(Orchestration Layer)
LLM Gateway
(openaiGpt5)
SNOW MCP
(ServiceNow Operations
MCP Server)
MuleSoft Platform MCP
(Mule Operations MCP
Server)
Traffic to the External services
Reasoning
MCP Invocation
Inbound Traffic
Agent
Fabric
✓ Email Triggered when incident created in ServiceNow
✓ Notification Listener will pick up the email and invoke the Agent
Broker through A2A protocol
✓ Agent Broker make a use of LLM Gateway and call appropriate
MCP Server
Mule
Operations
Broker

Agent Network Project Structure & Component
Quick Look
An agent network project defines a structured configuration for multi-agent systems, enabling orchestration of AI agents with
external services, tools, and inter-agent communication. This format provides a declarative way to define agent capabilities,
dependencies, and service integrations.
Project
Structure
✓ exchange.json: Contains asset metadata available in Anypoint
Exchange after publishing your agent network assets. Variable
layer - secrets, metadata, URLS.
✓ agent-network.yaml: Contains agent broker, agents, MCP, LLM
details and connections. Asset Registry and Gateway binding
✓ .agent files: Contain broker and node definitions and
configurations (instructions) that enable multi-agent
orchestration in your project. Dynamic Routing.

Quick Look
Project
Structure
Guided Determinism Pattern
2. Deterministic (rule-based nodes)
✓ Handle predefined, guaranteed execution paths.
1. Probabilistic (LLM-based nodes)
✓ Handle open-ended, judgment-heavy tasks.
Subagent Orchestrator Generator Executor
Uses an LLM Yes (reasoning loop) Yes (reasoning loop) Yes (single call) No
Calls actions
(A2A/MCP)
Yes—LLM chooses when
and how
Yes—LLM
coordinates multiple
agents/tools
No Yes—deterministic,
fully-bound arguments
Number of LLM
iterations
Multiple (configurable loop) Multiple (configurable
loop)
Exactly one Zero
Slot filling (…) Yes Yes No No
Structured
output schema
Yes Yes Yes No
Best for Single-task reasoning:
classification, extraction,
semantic routing, human-
in-the-loop decisions
Multi-agent
coordination:
delegating to several
agents/tools to
achieve a complex
goal
One-shot text
generation:
summaries,
formatting, templated
responses
Deterministic
execution: calling
actions with known
arguments, setting
variables, no LLM
judgment needed
Node
Types
Guided Determinism separates LLM-based probabilistic decisions from deterministic control-flow, combining AI judgment with predictable, rule-based execution.

Quick Look
Project
Structure

Core Building Blocks of Agent Fabric
Quick Look
Enterprise Actionability is the foundation that makes your existing APIs, data, LLMs, and services directly usable inside agentic
workflows — bridging today’s systems with the agentic world, without rebuilding.
MCP Bridge enables you to create MCP servers from your existing API instances, exposing their operations as MCP tools that agents
can discover and call, without modifying the underlying API.
MCP Bridge
• POST /api/appointment
• GET /api/appointment
Appointment Booking API
Omni
Gateway
1. Publish
API to
Exchange
2. Create
MCP Bridge
3. Configure
the
Upstream
Service
4. Map API
Resources
to MCP
Tools
5. Review &
Deploy
6. Secure &
Monitor
TOOLS : -
✓ get_appointment
✓ book_Appointment
Agent Ready MCP Server
LLM
Agent
From Rest API to MCP Tool – Fast, Seamless, Agent-Ready.

Quick Look
Anypoint Connector for MCP (MCP Connector) lets you build Mule applications that act as MCP servers and clients, giving you full
programmatic control over what tools and resources you expose to agents and enabling Mule apps to connect to external MCP
servers.
MCP Connector

Use Anypoint Connector for A2A (A2A Connector) to add A2A protocol support to Mule applications so they can act as both A2A
servers and A2A clients. You can also use A2A Connector to make agents built on other platforms A2A-compliant.
A2A Connector
AI Connectors provide pre-built integrations to LLMs and vector databases, making them directly accessible inside Mule-based
agentic workflows.
Other AI Connectors
Anypoint Code Builder is the development environment for building agent networks, configuring agent interactions, and developing
custom MCP servers.
Anypoint Code Builder
MuleSoft Intelligent Document Processing (IDP) uses multimodal large language models to extract structured data from unstructured
or semi-structured documents like invoices, purchase orders, receipts, and custom document types.
IDP
MuleSoft Vibes is the AI assistant built into Anypoint Code Builder, trained on industry best practices for graph-based agent
architecture.
MuleSoft Vibes
Quick Look

Prerequisites
Anypoint Platform
Component
Required Permissions
Runtime Manager ✓ Manage Servers
✓ Create Applications
✓ Read Servers
Anypoint Code
Builder
✓ Anypoint Code Builder
✓ Mule Developer
Generative AI User
Anypoint Design
Center
✓ Design Center Developer
API Manager ✓ Deploy API Proxies
✓ Manage Policies
Exchange ✓ Exchange Contributor
Usage ✓ Usage Viewer
Infrastructure ✓ CloudHub 2.0
Anypoint Platform
Environment Setup
✓ ACB Updated Version (1.23+)
✓ Agent Force Enabled Org
✓ Salesforce Tenant Relationship Established
✓ Omni Gateway Configured

Demo
Setup Agent Network Gateway (Managed Omni)
Create Agent Network Project
Publish Agent Network Asset to Exchange
Deploy Agent Network
Test

Flex Gateway – Latest Updates
• MuleSoft has rebranded Flex Gateway to Omni Gateway and broadened its scope from a
pure API gateway to a unified agent control-plane and governance layer for APIs, MCP
servers, agents, and LLMs.
Omni Gateway the evolved control plane for APIs, MCP servers,
agents, and LLMs
Old (Flex Gateway) New (Omni Gateway) Changed?
Flex Gateway Self-Managed Omni Gateway Renamed
Connected Mode Connected Mode Same
Local Mode Local Mode Same
Managed Flex Gateway Managed Omni Gateway Renamed
API Manager policies API Manager policies Same

Pune Meetup - MuleSoft Agent Fabric Inside

More Related Content

Similar to Pune Meetup - MuleSoft Agent Fabric Inside

Recently uploaded

Pune Meetup - MuleSoft Agent Fabric Inside