The art of simplicity

This final post continues on the multi-agent path: instead of one agent doing more things, we compose multiple agents doing the right things. Yesterday I demonstrated how to do this inside the Copilot SDK itself, today we look the broader ecosystem and I’ll show you how to integrate your Copilot SDK agent in the Microsoft Agent Framework. Microsoft Agent Framework The Microsoft Agent Framework(MAF) is the unified successor to Semantic Kernel and AutoGen. It provides a standard interface for building, orchestrating, and deploying AI agents. Dedicated integration packages let you wrap a Copilot SDK client as a first-class MAF agent — interchangeable with any other agent provider in the framework. The key distinction from Part 1: custom agents inside the SDK work within a single Copilot session, with the Copilot runtime as orchestrator. MAF operates at a higher level — it can compose a Copilot SDK agent with agents backed by Azure OpenAI, Anthropic, or any other provider, using st...

We've covered a lot of ground in this series: sessions and lifecycle, deployment and scaling, MCP integration, and skills. Each post added a new capability layer to a single agent. This post changes the shape of the problem: instead of one agent doing more things, we compose multiple agents doing the right things. There are two different levels at which you can do this. Inside the Copilot SDK itself, custom agents let the Copilot runtime orchestrate specialised sub-agents automatically within a single session. Beyond the SDK, the Microsoft Agent Framework lets you compose Copilot SDK agents with agents from any other provider in structured multi-agent workflows. In this post we stay inside the SDK. Our next and final post will look at the broader ecosystem and Microsoft Agent Framework integration. The problem with a single agent A single session with a broad system prompt is fine for many use cases. But as tasks grow more complex, the cracks show: A general agent need...

Up to this point in the series, we've extended the agent's capabilities in two ways: custom C# tools that execute code, and MCP servers that connect to external systems. Both are about what the agent can do . Skills are about what the agent knows and more specifically, how to apply that knowledge in a repeatable, composable way. A skill is a folder containing a SKILL.md file with specialised instructions, and optionally supporting files like scripts, templates, or examples. You point the SDK at a directory, the agent scans for skills automatically, and when a task matches a skill's description, that skill's instructions are injected into the session context. The agent gains focused expertise — without you writing any C#, and without redeploying your application. Why skills exist Think about a development task you repeat across projects: scaffolding a solution, writing a particular kind of test, generating API documentation in a house style, following a specific ...

After a short break, I'm back with my series about the GitHub Copilot SDK. So far, we have covered: You don't need to build your own agent harness Getting started with the GitHub Copilot SDK in .NET Sessions in the GitHub Copilot SDK: What they are and how to manage them Deploying and scaling Authentication and observability Today, we cover MCP integration — connecting your agent to external context and services via the Model Context Protocol, so your agent can reach databases, APIs, and cloud services without you having to build the glue. The agent we've built so far knows how to reason and plan, and can call C# tools you define directly. But there's a whole ecosystem of pre-built capabilities out there. Servers that talk to Azure, GitHub, databases, Slack, internal systems — all speaking the same open standard. MCP is how your agent connects to that ecosystem. Instead of writing tool integrations from scratch, you point the SDK at an MCP serv...

In 2012(!) I blogged about rubber duck debugging to help me troubleshoot issues. It's a practice I'm still using even today. With GitHub Copilot CLI's newest update, rubber duckis becoming AI driven. Copilot introduces a second AI model from a different family to critique your agent's plans and implementations at the moments where feedback has the highest return. Let's have a look at it in more detail... What rubber duck actually does Rubber Duck is not a general-purpose chat assistant. It's a review agent that will use a model from a complementary AI family to whichever model you've selected as your orchestrator. When you're running a Claude model as your orchestrator, Rubber Duck runs on GPT-5.4. This is intentional: a model reviewing its own output is still bounded by its own training biases. A cross-family reviewer brings genuinely different blind spots to the table. Rubber Duck's output is deliberately narrow — a short, prioritised lis...

Having your AI agent come up with a plan before start coding, has helped me a lot in my agentic development workflow. It allows me to verify the steps the agent will take and allowed me to avoid that the agent goes down the wrong path. To help you with this, Visual Studio got a ‘plan mode’ that once enabled allowed the agent to create a plan. I really liked the feature, the only problem is that is what not always obvious when the agent decides to create a plan and when it just starts executing. To tackle this issue, the plan mode in Visual Studio has evolved to a separate plan agent, similar to what we have in VSCode. Plan first, code second If you have never heard about plan mode or the plan agent, it is a dedicated mode in Copilot Chat that focuses entirely on understanding what you want to build before touching a single file. Instead of jumping straight to implementation, it asks clarifying questions, reads your codebase using read-only tools, and drafts a detailed implemen...