You don’t need code to build your first AI agent
Microsoft Developer demonstrates how to build a no-code AI agent in the Azure AI Foundry Portal: create a Foundry project, deploy a model, connect an MCP server for a cupcake store scenario, then validate the agent using tracing, token usage metrics, and evaluations.
Overview
This video walks through building the “Sparkles” agent using the Azure AI Foundry Portal, focusing on end-to-end setup (project + model + agent), connecting external tools via MCP, and using built-in observability and evaluation features to move toward production-ready agents.
Key links:
- Foundry Portal: https://aka.ms/foundry-portal
- Playlist: https://aka.ms/InsideMicrosoftFoundryPlaylist
Steps shown in the portal
Explore the Microsoft Foundry Portal
- Open the Foundry Portal experience and review the main navigation for projects, models, agents, tracing, and evaluations.
Create a project
- Sign into Foundry.
- Create a new Foundry project to hold the agent configuration and related resources.
Find and deploy a model
- Browse available models.
- Deploy GPT-5.4 Mini for use by the agent.
Create an agent (prompt agent)
- Create a new agent as a Prompt Agent.
- Configure the agent’s persona for the cupcake shop scenario (the “Sparkles” agent).
Connect an MCP server
- Connect the agent to the Cupcake Store MCP server.
- Use the MCP connection so the agent can:
- List available cupcake flavors
- Help customers place orders
Test the cupcake ordering agent
- Run an interactive test of the agent to confirm it can use the MCP server to complete the ordering flow.
Observability and quality checks
Inspect agent traces
- Use Tracing in the portal to review the agent’s execution traces.
- Inspect how the agent handled requests and tool calls.
Measure performance and token usage
- Review performance signals and token usage to understand cost/efficiency characteristics.
Create and run evaluations
- Create an Agent Evaluation from the portal.
- Select evaluation criteria.
- Run the evaluation and review the results.
Production readiness
- The video closes with guidance on using tracing and evaluations as part of building more production-ready agents.
Video chapters
- 0:00 – Explore the Microsoft Foundry Portal
- 0:24 – Create Project | Sign into Foundry and Create a New Foundry Project
- 1:12 – Navigate the Foundry Experience
- 2:12 – Models | Find and Deploy GPT-5.4 Mini
- 3:14 – Create Agent | Build a Prompt Agent
- 4:28 – Connect an MCP Server
- 5:40 – Test the Cupcake Ordering Agent
- 7:04 – Tracing | Inspect Agent Traces
- 7:53 – Measure Performance and Token Usage
- 8:10 – Evaluations | Create an Agent Evaluation
- 9:31 – Select Evaluation Criteria
- 10:03 – Review Evaluation Results
- 10:34 – Building Production-Ready Agents
- 10:52 – Summary