2026

Mar 18

Visualizing and checkpointing workflows in Microsoft Agent Framework

Microsoft Agent Framework Workflows 📚 Microsoft Agent Framework

A workflow is a graph that runs in discrete supersteps. That structure makes two things easy that are usually hard: drawing the graph, and pausing the run cleanly so you can resume it later. This article looks at WorkflowViz for visualization and the CheckpointStorage system for persistence.

Mar 16

Workflow fundamentals in Microsoft Agent Framework

Microsoft Agent Framework Workflows 📚 Microsoft Agent Framework

Sequential, concurrent, handoff, group chat, and magentic workflows are all built on the same four primitives: executors, edges, events, and a superstep-based execution model. This article looks at those primitives directly, so you know what is happening underneath when you reach for an orchestration builder, and so you can build workflows the builders do not cover.

Mar 15

Context providers and pluggable memory in Microsoft Agent Framework

Microsoft Agent Framework Agents Memory 📚 Microsoft Agent Framework

AgentSession keeps one conversation alive for as long as the Python process runs. Context providers let an agent remember things across sessions, across users, and across long timescales. This article looks at the provider abstraction, the built-in history and memory providers, and how to write your own.

Mar 10

Exposing Microsoft Agent Framework agents as MCP servers

Microsoft Agent Framework Tools MCP 📚 Microsoft Agent Framework

We have spent the last two articles consuming MCP servers from a MAF agent. This one flips the picture: how to publish a MAF agent as an MCP server so that other agents and applications, in any framework, can call it as a tool.

Mar 9

Hosted MCP tools in Microsoft Agent Framework

Microsoft Agent Framework Tools MCP Foundry 📚 Microsoft Agent Framework

Foundry can host the MCP connection on the agent's behalf. This article looks at how a hosted MCP tool differs from the local MCP setup we built last time, where authentication moves, and how to wire one up via FoundryChatClient.get_mcp_tool.

Mar 8

Local MCP tools in Microsoft Agent Framework

Microsoft Agent Framework Tools MCP 📚 Microsoft Agent Framework

Function tools cover code you write yourself. MCP servers cover tools that other people write. This article looks at how to connect a Microsoft Agent Framework agent to a local MCP server, what the agent actually sees when you do, and how MCP tools coexist with the function tools we have already used.

Mar 6

Structured outputs in Microsoft Agent Framework

Microsoft Agent Framework Agents 📚 Microsoft Agent Framework

Free-text replies are fragile to parse, especially when downstream code depends on the answer's shape. This article looks at how Microsoft Agent Framework constrains an agent's reply to a Pydantic model, when to use it, and how it interacts with tool calls and streaming.

Mar 5

Middleware in Microsoft Agent Framework

Microsoft Agent Framework Middleware Production 📚 Microsoft Agent Framework

MAF middleware lets you intercept agent runs, tool calls, and the underlying chat-client requests. This article looks at the three middleware types, when each one fires, and how to use them for logging, redaction, blocking, and result overrides without touching your agent code.

Mar 4

Streaming and multi-turn conversations in Microsoft Agent Framework

Microsoft Agent Framework Agents 📚 Microsoft Agent Framework

Real applications rarely call an agent once and walk away. They hold conversations, and they often stream the response as it is generated. This article looks at AgentSession, streaming responses, and how the two work together.

Mar 2

Function tools in Microsoft Agent Framework

Microsoft Agent Framework Agents Tools 📚 Microsoft Agent Framework

Function tools turn ordinary Python functions into capabilities your agent can call. This article looks at how MAF converts type hints into schemas, what return types it accepts, how async tools work, and what happens when a tool raises an error.

Mar 1

Choosing the right Microsoft Agent Framework client

Microsoft Agent Framework Agents 📚 Microsoft Agent Framework

Across this series, we have used several Microsoft Agent Framework client classes to talk to different model surfaces. This article maps them out, explains when each one fits, and shows the same simple agent built three different ways.

Feb 28

Microsoft Agent Framework - Agent Skills

Microsoft Agent Framework Workflows Agents Skills 📚 Microsoft Agent Framework

Latest release of Microsoft Agent Framework offers support for Agent Skills. We will explore implementing a practical use case for Skills with MAF.

Feb 28

Google ADK - Agent Skills

Google Agent Development Kit Agents Skills 📚 Google ADK

Dive into a powerful new feature in Google ADK: Agent Skills. We will explore what Agent Skills are, why they matter, and how to implement a practical use case.

Feb 15

GitHub Agentic Workflows

GitHub Agentic Workflows Agents 📚 GitHub Agentic Workflows

GitHub Agentic Workflows let you write automation in natural language. Instead of stitching together YAML steps, shell scripts, and third-party Actions, you describe what you want an AI agent to accomplish and give it the tools to do it.

Feb 6

Agent skills vs Model Context Protocol - [How] do you choose?

Agent Skills MCP 📚 Agents

Agent Skills and MCP are a hot topic. Understanding these standards genuinely overlap, where they serve fundamentally different purposes, and where each falls short, is essential for anyone building agent-powered systems.

2025

Dec 23

Google ADK - OpenAPI tools, agents-as-tools, authentication, and long-running operations

Google Agent Development Kit Agents 📚 Google ADK

Your agent needs to talk to the outside world. It needs to call REST APIs, delegate complex work to specialized agents, authenticate with protected services, and kick off operations that take minutes instead of milliseconds. Google ADK provides four powerful patterns to make all of this possible.

Dec 20

Google ADK - MCP tools

Google Agent Development Kit Agents 📚 Google ADK

An agent that can only use the tools you hard-code into it has a ceiling. The Model Context Protocol (MCP) shatters that ceiling by giving your ADK agents a standardized way to discover and use tools hosted anywhere — local processes, remote servers, or cloud services.

Dec 18

Google ADK - Runner and execution architecture

Google Agent Development Kit Agents 📚 Google ADK

You have agents, tools, callbacks, and sessions. But what actually runs them? The Runner is the central orchestrator that powers every agent interaction in Google ADK, driving the event loop that connects all these pieces together.

Dec 12

Google ADK - Callbacks

Google Agent Development Kit Agents 📚 Google ADK

Callbacks are user-defined functions that hook into an agent's execution pipeline at predefined checkpoints. They let you observe, customize, and control agent behavior without modifying the ADK framework code.

Dec 5

Google ADK - Sessions, state, and memory

Google Agent Development Kit Agents 📚 Google ADK

LLMs are stateless. Every API call to an LLM is independent. The model does not inherently remember what was said before. Yet meaningful conversations are inherently multi-turn, contextual, and stateful.

Dec 3

Google ADK - types of agents

Google Agent Development Kit Agents 📚 Google ADK

Google Agent Development Kit (ADK) is a flexible and modular framework for developing and deploying AI agents. It offers different types agents.

Oct 20

Magentic workflows in Microsoft Agent Framework

Microsoft Agent Framework Workflows Agents 📚 Microsoft Agent Framework

If you need a workflow that can adapt its routing decisions based on what agents discover during execution, look no further. Magentic is for you!

Oct 18

Group chat workflows in Microsoft Agent Framework

Microsoft Agent Framework Workflows Agents 📚 Microsoft Agent Framework

Building dynamic multi-agent discussions where AI agents collaborate, debate, and converge on solutions with GroupChatBuilder

Oct 16

Handoff workflows in Microsoft Agent Framework

Microsoft Agent Framework Workflows Agents 📚 Microsoft Agent Framework

Learn how to orchestrate AI agents in a concurrent workflow using Microsoft Agent Framework's HandoffBuilder

Oct 16

Docker cagent - An introduction

Docker cagent Agents 📚 Docker cagent

Docker cagent is an open-source, multi-agent AI runtime that lets you build, orchestrate, and share teams of specialized AI agents — all defined declaratively in YAML.

Oct 10

Building persistent AI Agents with Microsoft Agent Framework and Microsoft Foundry

Microsoft Agent Framework Foundry Agents 📚 Microsoft Agent Framework

The Microsoft Agent Framework represents a significant improvement in developer experience over the raw Foundry SDK. While the native SDK gives you complete control, the Agent Framework provides less boilerplate, type-safe tools, consistent patterns, and resource safety.

Oct 4

Building AI Agents with Microsoft Agent Framework

Microsoft Agent Framework Agents 📚 Microsoft Agent Framework

The world of AI agents is evolving rapidly, and Microsoft's Agent Framework provides a powerful, unified foundation for building intelligent agents that can reason, take actions, and interact with users naturally. In this blog post, I'll walk you through a series of practical examples that demonstrate how to create and use Azure AI agents using MAF and Python.

Oct 3

Getting Started with Foundry Agents

Microsoft Foundry Agents 📚 Microsoft Foundry

Microsoft Foundry (formerly Azure AI Foundry) is a unified system for building intelligent agents. It is a platform that provides models, tools, frameworks, and other aspects such as observability, guardrails, and enterprise-ready governance for creating AI agents.

Oct 2

Introduction to Microsoft Agent Framework

Microsoft Agent Framework Agents 📚 Microsoft Agent Framework

Microsoft Agent Framework is an open-source development kit for building AI agents and multi-agent workflows.

Sep 8

Azure AI Search - Vectorization, Custom Skills, and RAG

Azure AI Search 📚 Azure AI Search

In this article, we cross into vector territory. We will chunk documents using the Text Split skill, generate embeddings with Azure OpenAI, store them in a vector-enabled index, and run both vector and hybrid queries.

Sep 2

Azure AI Search - Getting Started

Azure AI Search 📚 Azure AI Search

Creating an enterprise AI application? Does your application need access to enterprise data and web content? Azure AI Search is the answer.

Jul 17

Local model serving - Using Foundry Local

inferencing local serving 📚 Local LLM Serving

There are several options available for running Large Language Model (LLM) inference locally. Foundry Local by Microsoft is a new entrant.

Jul 16

Local model serving - Using Docker model runner

inferencing local serving 📚 Local LLM Serving

Docker Model Runner — a faster, simpler way to run and test AI models locally, right from your existing workflow. Whether you’re experimenting with the latest LLMs or deploying to production, Model Runner brings the performance and control you need, without the friction.

Jul 15

Local model serving - Using LM Studio

inferencing local serving 📚 Local LLM Serving

There are several options available for running Large Language Model (LLM) inference locally. LM Studio is one such option. It is more comprehensive and offers some great features.

Jul 14

Local model serving - Using Ollama

inferencing local serving 📚 Local LLM Serving

There are several options available for running Large Language Model (LLM) inference locally. Ollama is one such option and my favorite among all. Ollama offers access to a wide range of models and has recently enabled cloud-hosted models as well. It offers both CLI and GUI (chat interface) to interact with the loaded models.

Apr 6

Changes in Autogen release 0.5.1

autogen azure-openai python 📚 AutoGen

Know what went into AutoGen 0.5.1 -- the open framework for creating multi-agent systems.

2024

Feb 3

Devcontainers for cloud-native application development

📚 Containers

Development containers are a great way to develop modern applications. Cloud-native applications usually implement more than one service to provide the application functionality. Using dev containers for microservices-based application development requires more than one container. This is where using Docker Compose with dev containers is useful. This article explores creating a multi-container development environment using VS Code dev containers and Docker Compose.

2023

2022

2021

2020

2019

2016

2015

2014

2013

2011

2010