Agent 365 SDK samples and prompts

Choose your starting point

Use a sample if you want to learn with an existing project before changing your own agent. If you already have an agent, you don't need to replace it with a sample:

Your goal	Start here	First result
Connect your existing Python, Node.js, or .NET agent with an AI coding assistant.	Connect an existing agent to Agent 365. Check its prerequisites first.	Register your agent, add selected capabilities, and validate the integration.
Register an agent blueprint by running the CLI yourself.	Set up an agent blueprint, starting with its prerequisites.	Create a blueprint for your agent's identity and permissions.
Learn with a sample project.	Choose your language in the selector, then use the sample code links.	Configure a sample, start it locally, and get a response in Agents Playground.

Sample code

Choose the framework you use or want to learn. Python and Node.js have step-by-step quickstarts; .NET samples start from a GitHub README. All sample source is in the Agent365-Samples repository.

Before running a sample, read the quickstart or README prerequisites for the runtime, model credentials, editor tooling, and authentication configuration. Use the testing guide for shared setup. Your first checkpoint is a response in Agents Playground, not a deployed agent or a successful live Microsoft 365 tool call.

Important

Local testing doesn't establish that tenant-backed capabilities are configured. Before testing live Microsoft 365 data, notifications, or telemetry, check the tenant access, permissions, licensing, and preview requirements for the capabilities you choose.

For a guided start, use the Python + Microsoft Agent Framework quickstart. Start with its prerequisites for Python, uv, and Agents Toolkit, and configure the sample's model credentials before testing.

For the first result, follow the quickstart's setup steps, then use Test your agent to configure a local Agents Playground conversation.

Sample	Description
Agent Framework Sample Agent	An example agent using Python and Microsoft Agent Framework.
Autonomous agent	An autonomous Python agent that summarizes trending GitHub repositories with Azure OpenAI and Agent 365 tracing.
Claude sample agent	An example Python agent using the Claude Agent SDK with extended thinking, built-in tools, and Agent 365 integration.
CrewAI agent sample	An example CrewAI multi-agent workflow for weather and driving safety with Agent 365 tracing, notifications, and MCP tools.
Google ADK sample agent	An example Python agent using Google ADK with Agent 365 observability, notifications, and MCP tools.
Observability with Azure Monitor	An example that adds Agent 365 observability to a Python app that already exports OpenTelemetry spans to Azure Monitor.
Observability with LangGraph	An example that adds Agent 365 observability to a Python LangGraph agent while preserving its existing OpenTelemetry exporter.
Observability with manual OpenTelemetry instrumentation	An example that manually instruments a Python agent with OpenTelemetry and Agent 365 tracing.
OpenAI Sample Agent	An example agent using Python and OpenAI Agents SDK.

Choose a guided quickstart based on your framework and model access:

Framework	Start here	Check first
LangChain	Node.js + LangChain quickstart.	Prerequisites for Node.js and Agents Toolkit, plus the sample README's Azure OpenAI or OpenAI credential setup.
Claude Agent SDK	Node.js + Claude Agent SDK quickstart.	Prerequisites for Node.js, Agents Toolkit, and Anthropic API credentials.

For either quickstart, the first checkpoint is a response from the configured sample in Agents Playground. Follow the quickstart's testing links for authentication and environment setup before sending a message.

Explore the sample READMEs for these and other Node.js frameworks:

Sample	Description
Autonomous agent	An autonomous TypeScript agent that summarizes trending GitHub repositories with Azure OpenAI and Agent 365 tracing.
Claude Sample Agent	An example agent using JavaScript and Claude Agent SDK.
Copilot Studio sample agent	An example that integrates a Microsoft Copilot Studio agent with Agent 365 using TypeScript hosting, notifications, and observability.
LangChain Sample Agent	An example agent using JavaScript and LangChain.js.
Devin Sample Agent	An example agent using JavaScript and Devin API.
n8n Sample Agent	An example agent using JavaScript and n8n.
OpenAI sample agent	An example OpenAI agent using TypeScript with Agent 365 observability, notifications, and MCP tools.
Perplexity Sample Agent	An example agent using JavaScript and Perplexity SDK.
Vercel Sample Agent	An example agent using JavaScript and Vercel AI SDK.

Start with the Semantic Kernel sample README. Check its Prerequisites section for the required .NET SDK and Semantic Kernel versions and Azure OpenAI or OpenAI API credentials. Follow the README's Running the Agent link to the testing guide to configure authentication, start the agent, and send a message in Agents Playground.

Sample	Description
Agent Framework simple sample	An example .NET agent using Microsoft Agent Framework with Agent 365 observability, notifications, and MCP tools.
Autonomous agent	An autonomous .NET agent that summarizes trending GitHub repositories with Azure OpenAI and Agent 365 tracing.
Semantic Kernel Sample Agent	An example agent using .NET and Microsoft Semantic Kernel.
W365 computer use sample	An example .NET agent that controls a Windows 365 Cloud PC using the OpenAI Responses API and W365 Computer Use MCP tools.

After your first response

Use the SDK validation checklist before deployment. When your scenario needs a hosted agent, continue with the deployment, publishing, instance verification, and cleanup guides. For recovery, use the testing troubleshooting steps or Agent 365 troubleshooting guide.

Prompts

Use Cursor IDE Prompt Guide to generate Agent 365 agents via natural language prompts using Cursor IDE.