Tech

Microsoft Ships Agent Framework 1.0 — Semantic Kernel and AutoGen Finally Merge

Microsoft's Agent Framework 1.0 merges Semantic Kernel and AutoGen into one production SDK for .NET and Python — with A2A and MCP interop, OpenTelemetry baked in, and an LTS commitment. The "which agent framework do we standardize on?" meeting just got shorter.

Initial Editor·2026-04-21·3min read·469 words

On April 3, Microsoft shipped Agent Framework 1.0 — the long-rumored merger of Semantic Kernel and AutoGen into one open-source SDK. If you've been juggling both (orchestration in AutoGen, enterprise plumbing in SK), this is the release that makes picking one unnecessary.

What's in the box

  • Single SDK for .NET and Python with matching primitives on both sides: Agent, AgentGroup, Tool, Workflow, Runtime.
  • Multi-provider model support — OpenAI, Azure OpenAI, Anthropic, local (Ollama, vLLM), and Azure AI Foundry under a common contract.
  • Cross-runtime interop via A2A and MCP. Agents built in Agent Framework can call — and be called by — LangGraph, CrewAI, OpenAI's Agents SDK, and any MCP-compliant server. Your "agent stack" choice is no longer a lock-in.
  • Multi-agent orchestration primitives: group chat, sequential handoff, planner/executor, and human-in-the-loop checkpoints — first-class, not recipe-book.
  • Observability out of the box: OpenTelemetry traces for every tool call, model call, and handoff.

What this actually changes

AutoGen was research-shaped — great for prototyping, rough in prod. Semantic Kernel was enterprise-shaped — great for integration, weaker on multi-agent ergonomics. 1.0 picks the winning half of each:

Kept from AutoGen Kept from SK Gone
Agent conversation model Plugin/connector ecosystem Two divergent ChatMessage types
Group-chat orchestration DI story, configuration Two streaming contracts
Planner/executor patterns Enterprise auth + secrets Framework-specific tool schemas

The interop story is the real news

A2A (Agent-to-Agent) and MCP (Model Context Protocol) being first-class means you can write a planner in Agent Framework that dispatches sub-tasks to a LangGraph worker running in someone else's cluster. Standardize on the protocols, not the framework.

Getting started

# .NET
dotnet add package Microsoft.AgentFramework --version 1.0.0

# Python
pip install agent-framework==1.0.0

Minimal Python example:

from agent_framework import Agent, Tool, Runtime

@Tool
def get_weather(city: str) -> str:
    return f"Sunny in {city}"

agent = Agent(
    model="gpt-5.4",
    tools=[get_weather],
    system="You are a concise travel assistant.",
)

with Runtime() as rt:
    print(rt.run(agent, "What's the weather in Hyderabad?"))

Migration notes

  • Semantic Kernel users get a compatibility shim — existing plugins keep working while you migrate orchestration code.
  • AutoGen users get a codemod that translates GroupChat / ConversableAgent into the new primitives.
  • LTS commitment: stable APIs, security fixes, no breaking changes on the 1.x line.

Should you switch?

If you're greenfield, yes — this is now the default answer for production .NET or Python agents. If you're mid-project on vanilla AutoGen or SK, the migration is mechanical enough that finishing the current milestone first and moving after is the right sequencing. Don't rewrite under deadline.

Sources

// more in tech

see all →
Tech· 2026-05-27· 5min

The Smallest Agent That Works, Part 2: The Three Reach-Out Agents

When the cheap tiers run out, the agent has to reach beyond the model itself — into knowledge it doesn't have, tools it can't natively use, or its own previous answer. RAG, tool use, and self-critique: three patterns, three failure modes worth pricing in. Part 2 of three.

#llm#rag#agent-architecture#ai-engineering
Tech· 2026-05-26· 5min

The Smallest Agent That Works, Part 1: The Three Cheap Agents

Most agent stacks are built one tier too capable for the job. Three of the cheapest architectures — a fixed pipeline, an LLM with rule constraints, and a reasoning loop — solve more problems than the architecture diagrams admit. Part 1 of three.

#llm#agent-architecture#ai-engineering#ai-agents
Tech· 2026-05-15· 5min

What MLX Got to Throw Away (That PyTorch Can't)

Every mature framework is a museum of decisions you can't take back. MLX is interesting mostly because it started after the decisions that matter for Apple Silicon were already mistakes — and the things it threw away are the things that were quietly costing the rest of us the most.

#ai-engineering#apple-silicon#mlx#ml-frameworks
Tech· 2026-05-15· 5min

The Unified-Memory Bet: Why On-Device Inference Stopped Being a Toy

For two years the industry's default answer to every inference question has been "bigger cluster." A different hardware topology is quietly making that the wrong default for a non-trivial slice of workloads — and the framework layer that earns it is the buzzword most decks haven't caught up with yet.

#hardware#ai-infrastructure#inference#edge-ai
Tech· 2026-05-14· 5min

Every Useful Skill Is One of Five Shapes

Skills aren't a freeform format. The useful ones fit one of five shapes — sequential workflow, multi-MCP coordination, iterative refinement, context-aware selection, domain-specific intelligence. Picking the right shape is most of the design work. Picking the wrong one is most of the bugs.

#claude-code#workflow#agents#skills
Tech· 2026-05-13· 5min

MCP Gives You the Kitchen. Skills Are the Recipe.

Most teams ship one of these and call the job done. MCP gives the agent tools. Skills tell it which to use, in what order, with which fallbacks. Without skills, your MCP integration ends with users asking 'okay, what now?'

#claude-code#mcp#agents#skills