Create Your First AI Agent from Scratch
In this session, I introduce **The Standard AI Agent Framework**, an open-source framework for building real, structured, observable, and extensible AI agents in C#. The framework is built on the Tri-Nature Theory of Agents: **Data:** What the agent has **Decision:** What the agent thinks **Direction:** What the agent does You will see how to build an agent, give it an identity through skills, connect it to tools, stream its activity in real time, and create a complete audit trail of its decisions and actions. In this video, we cover: • Building an AI agent in C# • Connecting to hosted or locally running language models • Programming agent behavior using natural-language skills • Giving agents access to local tools • Preventing unreliable LLM-generated calculations • Streaming agent activity, tool usage, status, and responses • Logging agent execution for auditability • Understanding gates, brains, judges, memory, knowledge, and external tools • Composing multiple agents together • Building the same framework in any programming language The Standard Agent architecture separates agent behavior into Data, Decision, and Direction, with dedicated foundations for skills, memory, knowledge, gating, reasoning, verification, internal tools, external tools, and final responses. An agent can even become a tool for another agent, allowing you to build larger multi-agent systems without introducing an entirely different architecture. **NuGet Package** https://www.nuget.org/packages/Standard.Agents **Source Code** https://github.com/hassanhabib/The-Standard-Agent **Language-Neutral Specification** Build a conformant implementation in Python, Java, Go, Rust, JavaScript, or any other programming language: https://github.com/hassanhabib/The-Standard-Agent-Specs/blob/main/SPEC.md The specification defines contracts and behavior rather than forcing a particular programming language or project structure, making the architecture portable across technology stacks. Try the framework, explore the source code, and let me know what kind of agent you build. 00:00 Introduction to The Standard Agent Framework 00:39 The Tri-Nature Theory of AI Agents 02:01 The Complete Agent Architecture 02:18 Data: Skills, Memory, and Knowledge 03:01 Decision: Gate, Brain, and Judge 03:35 Direction: Internal and External Tools 04:28 Building an AI Agent in C# 05:15 Creating Your First Standard Agent 06:04 Connecting to a Local LLM 06:50 Sending the First Prompt 07:42 Why an LLM Is Not Yet an Agent 07:58 Teaching the Agent New Skills 10:02 Configuring Custom Skill Directories 10:58 Programming Agent Identity and Behavior 11:06 Giving the Agent Tools 11:25 Building a Calculator Tool 12:19 Testing the Agent Without Tools 13:37 Teaching the Agent to Use Tools 14:27 Registering the Calculator Tool 15:25 The Agent Produces the Correct Answer 15:50 Logging and Auditing Agent Behavior 17:06 Streaming the Agent’s Thoughts and Actions 19:34 Monitoring Tool Execution 20:00 Tracking Agent Status and Responses 20:39 Memory, MCP Servers, Gates, and Judges 21:30 Multi-Agent Orchestration 22:07 Environment-Aware AI Agents 22:34 Why the Full Framework Will Be a Series 22:52 Building Standard Agents in Any Language 23:17 Final Thoughts
Build Reliable Event-Driven Systems Without Cloud Lock-In | EventHighway 2.0
Event-driven systems should be reliable, observable, and portable. They should not force your application into a specific cloud platform or require an external HTTP endpoint for every event handler. In this session, Christo du Toit introduces EventHighway 2.0, a Standard-compliant .NET library for building durable event-driven systems. EventHighway lets an event fan out to matching listeners while maintaining individual delivery records that can be observed, retried, archived, and replayed. Christo walks through the architecture and source code behind the new release, including: 0/ Durable and observable event delivery 1/ In-process event handlers with no HTTP endpoint required 2/ Event filtering and promoted properties 3/ Loop detection and event quarantine 4/ Participants and rotating secrets 5/ Failed and historical event replay 6/ SQL Server and PostgreSQL storage providers 7/ Health metrics and the Blazor operations portal 8/ Local, cloud-independent deployment EventHighway 2.0 delivers events directly to registered in-process handlers, while creating a separate delivery record for every matching listener. This allows each delivery to carry its own status, response, retry history, and replay state. The session also explores the idea of building cloud-foreign software: systems that can run on Azure, AWS, Google Cloud, private infrastructure, or entirely on a developer’s local machine. EventHighway is intentionally platform-agnostic, allowing events to move from anywhere to anywhere without tying the application to a particular cloud vendor. Source code https://github.com/The-Standard-Organization/EventHighway NuGet package https://www.nuget.org/packages/EventHighway The Standard https://github.com/hassanhabib/The-Standard EventHighway supports immediate and scheduled events, filtering, observability, configurable loop detection, participant authorization, retries with Fibonacci backoff, archiving, targeted replay, and pluggable storage providers. Replayable. Reliable. Local. HUGE thanks to the awesome Standard Community members starting with Christo Du Toit, Greg Hays, Zafar, Abdulsamad and Ahmad Salim you can chat with us here on Discord: https://discord.com/invite/vdPZ7hS52X 00:00 Introducing EventHighway 2.0 00:36 What Is EventHighway? 01:14 Pub/Sub and Event Addresses Explained 04:09 What Is New in Version 2? 05:14 Filtering Events at the Listener 06:29 Detecting and Quarantining Event Loops 07:13 Participants, Metrics, and Observability 07:45 Replaying Failed and Historical Events 09:22 Escaping Cloud Lock-In 10:54 Running Your Entire System Locally 12:06 EventHighway 2.0 Source Code Tour 13:38 The Basic Application Sample 14:41 Upgrading from EventHighway V1 16:49 Simulating APIs with WireMock 18:01 Configuring Loop Detection and Handlers 19:31 Building Custom In-Process Handlers 21:03 The Biggest Change from V1 to V2 22:36 Supporting Existing REST API Integrations 24:04 Delivery Results and Observability 25:27 Handling Credentials More Securely 26:27 The Risk of Third-Party Handler Packages 27:26 Sandboxing Custom Event Handlers 29:01 Registering Handlers, Participants, and Addresses 30:00 Participant-Based Security and Auditing 32:00 Filtering Events with Dynamic Expressions 33:00 Running the EventHighway Demo 34:00 Loop Detection and Event Quarantine 36:00 Correlation IDs and Future Traceability 37:00 Replaying Quarantined Events 38:00 Exploring the Blazor Operations Portal 39:00 Running EventHighway Completely Locally 40:00 Archiving and Bulk Event Replay 42:00 Participant Secrets and Secret Rotation 43:00 Identity, Users, and Participant Management 44:00 Self-Service Administration 45:00 The EventHighway Substrate Sample 46:00 Sending Events Through the REST API 47:00 Loop Detection in a Live Application 48:00 Resilient Retries with Fibonacci Backoff 50:00 Replay, Auditability, and Fresh Retry Budgets 50:48 Introducing the Event Substrate 52:00 Connecting Services Through Event Handlers 53:00 Persisting Data and Emitting Response Events 54:00 A New Receiver Pattern 54:27 The Future of Fully Event-Driven Systems 56:00 Traditional Layered Architecture 57:00 Observability Beyond Fire-and-Forget 58:00 Completely Detaching Software Modules 59:00 Zero-Downtime Service Replacement 60:00 Event-Driven MCP Servers and AI Tools 61:00 Why Dependency Injection Still Creates Coupling 62:00 Standard 3.0 and the Substrate Dependency 63:00 Building Durable, Cloud-Independent Services 64:00 Where the Hybrid Architecture Still Fits 65:00 Human-Comprehensible Software Architecture 66:00 Applying the Pattern to PeerLLM 67:00 Why AI Still Needs Good Software Architecture 68:00 The Standard Software License Promise 69:00 Shoutout to Christo Du Toit and Closing #dotnet #csharp #softwarearchitecture #eventdrivenarchitecture #pubsub #opensource #programming
Announcing .NET Modernization for Beginners
<p>A free, open-source, hands-on course that walks you through modernizing a real legacy ASP.NET application all the way to .NET 10 using the GitHub Copilot modernization agent, step by step.</p> <p>The post <a href="https://devblogs.microsoft.com/dotnet/announcing-dotnet-modernization-for-beginners/">Announcing .NET Modernization for Beginners</a> appeared first on <a href="https://devblogs.microsoft.com/dotnet">.NET Blog</a>.</p>
.NET AI Community Standup: What’s New: Agents, Local AI, and More! 🚀
Join Bruno Capuano and Pablo Lopes for a new .NET + AI Community StandUp featuring an informal and practical conversation about the latest developments across .NET and Artificial Intelligence. We’ll explore Microsoft Agent Framework, Agent Skills for .NET, Local AI, GitHub Copilot-powered application modernization, and more. Whether you’re already building intelligent .NET applications or just getting started, join us to discover new tools, ideas, and possibilities for your next AI project. ⌚ Chapters: 00:00 Countdown 02:00 Welcome and Introductions 04:40 Local AI with .NET overview 05:15 Why run AI on your machine 06:35 One interface. Many runtimes. - Microsoft.Extensions.AI 11:16 Ollama: local models, OpenAI compatible 12:30 Foundry Local: on-device, Microsoft-native 15:10 The Foundry Local lifecycle 19:00 Demo 28:36 More on local AI LLMs and tool calling 33:00 Foundry documentation 36:00 Agent Memory for .NET overview 42:12 neo4j overview 45:50 Getting Started with Agent Memory for .NET https://github.com/joslat/agent-memory-dotnet/ 48:25 Microsoft Agent with Memory documentation 52:30 Context Providers 58:56 Agent Memory sample session 01:01:20 Thanks for watching! 01:02:00 Wrap 🔗 Links: Foundry Local Blog: https://devblogs.microsoft.com/foundry/foundry-local-ga/ https://learn.microsoft.com/en-us/collections/8xymaztoogqjd?source=docs 🎙️ Featuring: Bruno Capuano (https://www.linkedin.com/in/elbruno/), Pablo Nunes Lopes (https://www.linkedin.com/in/pablonuneslopes/) #dotNET #ArtificialIntelligence #MicrosoftAgentFramework #LocalAI #GitHubCopilot #AzureAI
Blazor Community Standup: Generative UI with AG-UI & A2UI
Join us to learn how AI agents can dynamically create and adapt application experiences using AG-UI, A2UI, Microsoft Agent Framework, and Blazor. ⌚ Chapters: 00:00 Countdown (silent) 02:33 Welcome and Intros 04:38 What is Generative UI (Agentic UI) 08:15 Why do we need Generative UI 10:50 Generative UI types 17:39 Agentic UI in .NET Blazor Applications - architecture 22:30 Microsoft Agent Framework 25:20 AG-UI Protocol overview 25:47 Support for AG-UI in .NET 30:45 AG-UI - using an agent int he app (pseudocode) 32:36 A2UI: dynamically rendering UI from JSON sent from an agent 34:55 A2UI - sample component definition 36:12 Demo 42:56 Q&A 47:58 Resources & links 48:30 Key takeaways 56:17 Wrap 🔗 Links: https://www.theurlist.com/blazor-standup-20260714 🎙️ Featuring: Daniel Roth, Javier Calvarro Nelson, Maya Mateva, Stamo Gochev #dotnet #aspnetcore #blazor
.NET and .NET Framework July 2026 servicing releases updates
<p>A recap of the latest servicing updates for .NET and .NET Framework for July 2026.</p> <p>The post <a href="https://devblogs.microsoft.com/dotnet/dotnet-and-dotnet-framework-july-2026-servicing-updates/">.NET and .NET Framework July 2026 servicing releases updates</a> appeared first on <a href="https://devblogs.microsoft.com/dotnet">.NET Blog</a>.</p>
CoreCLR Progress and the Mono Timeline for .NET MAUI
<p>CoreCLR is now the only runtime for .NET MAUI mobile apps in .NET 11 Preview 6. Here is where we are, and why we need you to test today.</p> <p>The post <a href="https://devblogs.microsoft.com/dotnet/coreclr-progress-and-mono-timeline-dotnet-maui/">CoreCLR Progress and the Mono Timeline for .NET MAUI</a> appeared first on <a href="https://devblogs.microsoft.com/dotnet">.NET Blog</a>.</p>
.NET 11 Preview 6 is now available!
<p>Find out about the new features in .NET 11 Preview 6 across runtime, SDK, libraries, ASP.NET Core, .NET MAUI, C#, Entity Framework Core, F#, and container images.</p> <p>The post <a href="https://devblogs.microsoft.com/dotnet/dotnet-11-preview-6/">.NET 11 Preview 6 is now available!</a> appeared first on <a href="https://devblogs.microsoft.com/dotnet">.NET Blog</a>.</p>
.NET Data Community Standup: SQL Database Project Power Tools
Long time EF Core contributor ErikEJ will show us his SQL Database Project Power Tools - enhancing the developer experience for SQL Database Projects in Visual Studio and SSMS. 🎙️ Featuring: Jiri Cincura,Erik Ejlskov Jensen #sqlserver #ssms #databases #visualstudio
.NET MAUI Community Standup: MAUI UI July Recap
Join Matt Goldman, David Ortinau, and Gerald Versluis for a recap of MAUI UI July: Matt’s month-long community initiative packed with .NET MAUI UI content, inspiration, tips, and demos. We’ll look back at the highlights, community contributions, latest updates, improvements, and what’s next for building beautiful cross-platform apps with .NET MAUI. 🔗 Links: https://www.theurlist.com/maui-standup-august26 🎙️ Featuring: David Ortinau (@davidortinau), Gerald Versluis (@jfversluis), Matt Goldman (@mattgoldman) #dotnetmaui #mauiuijuly #dotnet
Don't let your AI Agents get hacked - watch this
How to make your AI Agents and MCP servers secure with the Microsoft Agent Governance Toolkit. Here we add Prompt Injection Prevention and Tool Calling Policies to an agent that uses the Microsoft Agent Framework. MVP Love code: https://github.com/edandersen/mvp-love Microsoft Agent Governance Toolkit: https://microsoft.github.io/agent-governance-toolkit/ 0:00 Intro 1:50 Microsoft Agent Governance Toolkit 2:47 Preventing unsafe tool calling 6:32 Stopping Prompt Injection 8:08 Securing MCP Servers with the same policies #coding #ai #dotnet