Houston Meetup Food Truck Rescue Agent Network

25-minute Houston MuleSoft Meetup Demo | May 2026

---

Demo Purpose

A meetup organizer needs to feed 80 attendees tonight with a $1,200 budget. This agent network handles the entire planning workflow — from venue constraints to dietary needs to final food truck recommendation — using a multi-agent broker pattern with MCP tool resources.

Repository Layout

File	Purpose
agent-network.yaml	Source of truth — broker card, skills, agents, MCP tools, LLM config
exchange.json	Anypoint Exchange asset metadata + runtime variable declarations
DEMO-CHECKLIST.md	Demo readiness checklist + 25-minute run of show + backup plan
.mvn/extensions.xml	Anypoint Code Builder Maven extension

---

Architecture

User Prompt
    │
    ▼
meetup-planning-broker  (OpenAI gpt-4o-mini)
    │
    ├── venue-logistics-mcp        ← venue rules, capacity, delivery windows
    ├── food-vendor-catalog-mcp    ← tacos, BBQ, Mediterranean, pizza, sandwiches
    └── budget-calculator-mcp      ← cost/person, total, tax/tip, fit check
    │
    ├── venue-policy-agent         (A2A)
    ├── attendee-preference-agent  (A2A)
    ├── food-options-agent         (A2A)
    ├── budget-check-agent         (A2A)
    └── recommendation-agent       (A2A)

---

Broker

meetup-planning-broker — orchestrates all agents in sequence and uses MCP resources as tools.

Agent coordination order:

1. venue-policy-agent — venue constraints (capacity, setup, delivery, cleanup)
2. attendee-preference-agent — dietary restrictions (vegetarian, vegan, gluten-free, halal)
3. food-options-agent — propose three food category options
4. budget-check-agent — validate options against $1,200 budget (~$15/person)
5. recommendation-agent — assemble final recommendation

Response format (six sections):

1. Recommended Plan
2. Why It Fits the Venue
3. Why It Fits the Budget
4. How Dietary Needs Were Handled
5. Agent Network Trace Summary
6. MCP Resources Used

---

Agents

Agent	Role
venue-policy-agent	Venue constraints: capacity, setup windows, loading, delivery, cleanup
attendee-preference-agent	Dietary needs across 80 attendees
food-options-agent	Food category proposals matching venue + dietary data
budget-check-agent	Budget validation at ~$15/person against $1,200
recommendation-agent	Final synthesized recommendation

All agents are A2A references. No complex agent logic is required for the demo.

---

MCP Resources

MCP Server	Purpose
venue-logistics-mcp	Room capacity, setup/cleanup windows, parking, loading dock, delivery time limits
food-vendor-catalog-mcp	Food categories and vendor-style options with dietary tag filtering
budget-calculator-mcp	Cost per person, total estimate, tax/tip assumptions, budget fit check

---

Test Prompt

Send this to the broker endpoint after deployment:

We have 80 people coming to the Houston MuleSoft Meetup tonight at CityCentre.
Budget is $1,200. We have vegetarians, a few gluten-free guests, and one halal requirement.
What food trucks should we get and will it fit the budget?

---

Runtime Variables to Replace

Variable	Description
ingressgw.url	Ingress gateway base URL (broker endpoint prefix)
openai.apiKey	OpenAI API key (secret)
venuePolicyAgent.url	A2A URL for Venue Policy Agent
attendeePreferenceAgent.url	A2A URL for Attendee Preference Agent
foodOptionsAgent.url	A2A URL for Food Options Agent
budgetCheckAgent.url	A2A URL for Budget Check Agent
recommendationAgent.url	A2A URL for Recommendation Agent
venueLogisticsMcp.url	Streamable HTTP URL for Venue Logistics MCP
foodVendorCatalogMcp.url	Streamable HTTP URL for Food Vendor Catalog MCP
budgetCalculatorMcp.url	Streamable HTTP URL for Budget Calculator MCP

---

25-Minute Run of Show

Time	Segment
0:00 – 2:00	Hook — "80 people, $1,200, food trucks in 4 hours. Let's solve this."
2:00 – 5:00	Problem framing — why multi-agent beats a single LLM call
5:00 – 9:00	Architecture walk — broker, 5 agents, 3 MCP servers, 1 LLM
9:00 – 13:00	Live file tour — agent-network.yaml sections: broker card, skills, links, tools
13:00 – 17:00	Demo run — send the test prompt, watch agent trace in real time
17:00 – 20:00	MCP resource spotlight — show budget-calculator-mcp output live
20:00 – 22:00	Exchange publish — publish_agent_network_assets live or pre-recorded
22:00 – 25:00	Q&A + closing

---

Next Step

Publish to Anypoint Exchange:

# Using Anypoint CLI
anypoint-cli-v4 exchange asset upload \
  --organization ffcd3f71-5bc6-4b00-b035-dceddd86fc80 \
  --version 1.0.0

Or use the publish_agent_network_assets MCP tool from Anypoint Code Builder.

---

Built with MuleSoft Agent Fabric · schemaVersion 1.0.0 · Houston MuleSoft Meetup May 2026