microsoft-foundry

microsoft-foundry / Foundry-Agent-Lab

Public

Ready-to-use structured and progressively complex agent demos with demo scripts and How-it works.

16
5
100% credibility
Found May 28, 2026 at 16 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Microsoft Foundry Agent Lab is an official educational project from Microsoft containing 9 progressive examples that teach you how to build AI agents. Starting from the simplest chat agent, each demo adds one new concept: function calling with live APIs, desktop and web interfaces, built-in web search with citations, code execution in a sandbox, document search (RAG), external tool integration via MCP, centralized tool governance with Toolbox, and self-hosted agents. The project uses Azure authentication (no hardcoded secrets), includes detailed presenter guides and demo scripts, and works with Microsoft Foundry's Model Router to automatically select the best AI model for each task.

How It Works

1
💡 You want to build an AI assistant

You've heard about AI agents and want to create one that can search the web, answer questions, and use tools.

2
📦 You download the learning kit

You grab a collection of 9 ready-to-run examples from Microsoft that teach you one concept at a time.

3
🤖 You create your first chat agent

With just a few lines of code, your agent comes to life and you can chat with it in the terminal.

4
🔧 You give your agent superpowers

You add abilities one by one: weather lookup, web search with citations, code execution, and searching through your own documents.

5
You choose your preferred interface
⌨️
Terminal chat

Type messages directly in your command window

🪟
Desktop window

Chat in a friendly pop-up window with scrollback

🌐
Web browser

Access your agent through any browser tab

6
🔒 You connect external tools safely

You link your agent to GitHub and other services with human approval required before any action runs.

🎉 You built a complete AI agent

Your agent can search the web, run code, read your documents, and connect to external services—all with enterprise-grade security.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 16 to 16 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is Foundry-Agent-Lab?

Foundry-Agent-Lab is a structured learning path for building AI agents with the Microsoft Foundry SDK in Python. It ships nine progressively complex demos that each introduce exactly one new concept, starting from a bare-bones terminal chat and ending with a self-hosted streaming agent. The demos cover function calling, built-in tools like web search and code interpreters, RAG with vector stores, MCP integration, centralized toolbox governance, and custom agent servers with the Responses protocol. Each demo runs via simple batch scripts and includes its own configuration, session logging, and cleanup scripts.

Why is it gaining traction?

The progressive structure is the hook: instead of dumping a full-featured agent on you, each demo adds one concept on top of the previous one. This makes it ideal for learning the Foundry SDK's mental model, especially the distinction between client-side function tools and server-side built-in tools. The Model Router feature is also a strong selling point, routing requests to the optimal model automatically without per-agent configuration. The demos demonstrate three different UX modes (terminal, Tkinter desktop, Gradio web) to show that the same agent works across presentation layers.

Who should use this?

Developers evaluating Microsoft Foundry for agent projects will get the most value here. If you need to present agent capabilities to stakeholders, the included demo scripts and presenter guide make it easy to walk through concepts in a structured way. Python developers comfortable with the OpenAI-compatible Responses API will find the patterns directly applicable. Teams already invested in Azure will appreciate the DefaultAzureCredential approach to authentication. This is less useful if you need production-ready patterns out of the box.

Verdict

At 16 stars with a 1.0% credibility score, this is an early-stage educational resource rather than a battle-tested library. The documentation is thorough and the demos are well-structured, but the low community engagement signals limited real-world validation. Use it as a learning tool and reference implementation, not as a dependency for production systems. If you're already on Azure and want to understand Foundry's agent capabilities, the progressive structure makes it worth your time.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.