At a glance
WHAT IT’S REALLY ABOUT
Hands-on workshop: deploy Claude in Foundry, build cupcake agent
- The session explains the shift from single-turn chatbots to agentic systems that plan, reason, and take actions across multiple steps and longer contexts.
- Microsoft Foundry is positioned as a unified, enterprise-ready platform for building and operating AI apps and agents with security, observability, governance, and broad tool connectivity.
- Attendees deploy and test a Claude model (Sonnet 4.6) in the Foundry playground, experimenting with system prompts to validate behavior.
- The workshop moves from playground to local development by configuring endpoint and API key settings, then running a simple agent using Microsoft Agent Framework in Python.
- An MCP (Model Context Protocol) server is added to provide tools, prompts, and resources—enabling the agent to query cupcake flavors, adopt a store persona, and complete an order workflow using vouchers and an approval dashboard.
IDEAS WORTH REMEMBERING
5 ideasAgentic apps introduce new requirements beyond model quality.
Multi-step reasoning over long context, enterprise reliability/observability/security, and tool/data connectivity are highlighted as core challenges that platforms must solve alongside better models.
Foundry’s value proposition is “prototype-to-production” for enterprise agents.
The talk emphasizes built-in security and governance plus integrations like Defender, Purview, and Entra ID so teams can avoid ad-hoc secret handling and scale deployments confidently.
Use the Foundry playground to quickly validate prompts and model behavior.
The demo shows swapping system prompts (serious vs. playful) to test responsiveness, tone, and instruction-following before embedding the model into an application.
Correct endpoint configuration is a common make-or-break step.
When moving to local dev, the endpoint must be set properly (not ending in “/v1/messages”); misconfiguration prevents the app from calling the hosted model successfully.
Microsoft Agent Framework provides a straightforward “hello world” path to an agent.
A small Python agent is created using environment variables (deployment/model, key, endpoint) and run locally to confirm the basic chat loop works before adding tools.
WORDS WORTH SAVING
5 quotesOver time or over the past few years especially, we've really seen the, the industry moving away from just single-turn AI conversations where you're just talking to an AI and it's giving you a response. And now we're actually building agentic systems where our agents can plan, um, they can reason, and they can take action on our behalf over time.
— Marlene Mhangami
This next wave of AI actually makes it so that even though we love the Claude models, and I think the models are getting better every day, you can't just rely on the models getting better. We also need systems to be able to actually have those models with that intelligence executed.
— Marlene Mhangami
Microsoft Foundry is our unified platform for building AI applications and agents at scale.
— Marlene Mhangami
We support over 1,400 built-in connectors and MCP tools, so your agents can take real action in the real world.
— Marlene Mhangami
MCP is Model Context Protocol, and this is an open standard for letting AI agents talk to external systems.
— Marlene Mhangami
High quality AI-generated summary created from speaker-labeled transcript.
