agent-framework-go

Welcome to Microsoft Agent Framework for Go!

Microsoft Agent Framework (MAF) is an open, multi-language framework for building production-grade AI agents and multi-agent workflows. This repository contains the Go implementation of Microsoft Agent Framework, along with Go samples and documentation to help you get started.

Microsoft Agent Framework is built for teams taking agents from prototype to production. It provides a consistent foundation for building, orchestrating, and operating agent systems while keeping architecture choices open as requirements evolve, and supports a broad ecosystem including Microsoft Foundry, Azure OpenAI, OpenAI, Model Context Protocol (MCP), Agent2Agent (A2A), AG-UI, and the GitHub Copilot SDK.

The .NET and Python implementations are available in the upstream Microsoft Agent Framework repository.

Note: Microsoft Agent Framework for Go is in public preview and is currently evolving outside the core upstream codebase. As adoption and feedback grow, we expect closer alignment with the broader MAF ecosystem.

Watch the full Agent Framework introduction (30 min)

Is this the right framework for you?

MAF is a strong fit if you:

• are building agents and workflows you expect to run in production,
• need orchestration beyond a single prompt or stateless chat loop,
• want graph-based patterns such as sequential, concurrent, group collaboration, and custom workflow routing,
• care about checkpointing, restartability, observability, governance, or human-in-the-loop control,
• need provider flexibility so your architecture can evolve without major rewrites.

Key Features

Explore new MAF capabilities and real implementation patterns on the official blog.

For a detailed .NET-to-Go feature comparison, see the .NET and Go SDK feature comparison. In short, the Go SDK covers core agents, tools, middleware, workflows, observability, and interoperability integrations, while .NET currently has broader product integrations and several features that are not implemented yet in Go.

• Go Support: Go packages, examples, and APIs for building agents and workflows in Go.
  • Go reference | Go examples
• Multiple Agent Provider Support: Support for various LLM and agent providers, with more being added continuously.
  • Provider examples | Provider packages
• Middleware: Flexible middleware for request/response processing, logging, OpenTelemetry, context providers, tool approval, and automatic tool calling.
  • Agent middleware | Agent harness
• Orchestration Patterns & Workflows: Build multi-agent systems with graph-based workflows supporting sequential, concurrent, group collaboration, conditional routing, subworkflows, checkpointing, streaming, human-in-the-loop, and time-travel patterns. Handoff orchestration is not implemented yet in the Go SDK.
  • Workflow examples | Workflow package
• Microsoft Foundry Agents: Build project-backed Foundry agents, invoke existing server-side Foundry agents, pass Foundry client headers, capture served-model metadata, and use Foundry memory. Foundry-hosted deployment is not implemented yet in the Go SDK.
  • Foundry examples | Foundry provider
• Observability: OpenTelemetry integration for distributed tracing, monitoring, and debugging.
  • Agent and embedding OpenTelemetry | Workflow OpenTelemetry
• Declarative Agents: Not implemented yet in the Go SDK.
• Agent Skills: Build domain-specific knowledge bases from files, inline definitions, and scripts for agents to discover and use.
  • Skills examples | Skills package
• Text Search (RAG): Retrieve and cite external knowledge automatically before agent runs or through an on-demand search tool, with optional session-backed recent-message memory.
  • Text search package | RAG examples
• Chat History Memory: Store conversations in a vector store and retrieve scoped memories automatically or through an on-demand tool.
  • Memory package | Memory examples
• Embeddings: Provider-independent batch generation, numeric and binary vectors, request options, usage metadata, and single-input/vector/batch-pair helpers, with an OpenAI/Foundry adapter and opt-in OpenTelemetry instrumentation.
  • Embedding package | OpenAI embedding generator | Foundry embedding generator
• Vector Data: Vector stores with typed and dynamic collections, CRUD helpers, and vector/hybrid search options with portable filters. Includes Azure AI Search and an in-memory connector with exact vector search, batch embedding generation, and JSON snapshots.
  • Vector data package | Azure AI Search | In-memory connector | Experimental provider services
• AF Labs: Not implemented yet in the Go SDK.
• DevUI: Not implemented yet in the Go SDK.

Table of Contents

• Welcome to Microsoft Agent Framework for Go!
  • Is this the right framework for you?
  • Key Features
  • Table of Contents
  • Getting Started
    • Installation
    • Learning Resources
    • Quickstart
      • Basic Agent - Go
  • More Examples & Samples
    • Go
  • Community & Feedback
  • Troubleshooting
    • Authentication
    • Environment Variables
  • Contributor Resources
  • Important Notes
    • Telemetry and data collection
    • Preview status
    • Trademarks

Getting Started

Installation

Go

go get github.com/microsoft/agent-framework-go

Learning Resources

• Overview - High level overview of the framework
• Quick Start - Get started with a simple agent
• Tutorials - Step by step tutorials
• User Guide - In-depth user guide for building agents and workflows

Quickstart

Basic Agent - Go

Create a simple Microsoft Foundry agent that writes a haiku about the Microsoft Agent Framework.

package main

import (
	"cmp"
	"context"
	"fmt"
	"os"

	"github.com/Azure/azure-sdk-for-go/sdk/azidentity"
	"github.com/microsoft/agent-framework-go/provider/foundryprovider"
)

func main() {
	endpoint := os.Getenv("FOUNDRY_PROJECT_ENDPOINT")
	model := cmp.Or(os.Getenv("FOUNDRY_MODEL"), "gpt-4o-mini")

	// Authenticate to Microsoft Foundry.
	token, err := azidentity.NewDefaultAzureCredential(nil)
	if err != nil {
		panic(err)
	}

	// Create a Microsoft Foundry agent.
	a := foundryprovider.NewAgent(endpoint, token, foundryprovider.ModelDeployment(model),
		foundryprovider.AgentConfig{
			Instructions: "You are a helpful assistant.",
		},
	)

	// Run the agent.
	ctx := context.Background()
	fmt.Println(a.RunText(ctx, "Write a haiku about the Microsoft Agent Framework").Collect())
}

More Examples & Samples

Go

• Getting Started: progressive tutorial from hello world to workflows
• Agent Concepts: deep-dive samples by topic, including tools, middleware, providers, observability, A2A, AG-UI, MCP, and skills
• File memory: persist per-user memories across separate conversations. Run go run ./examples/02-agents/agents/step24_filememory from the repository root with FOUNDRY_PROJECT_ENDPOINT and FOUNDRY_MODEL set, after authenticating with az login. Files remain in agent-memory/users/UID1 between runs.
• Agent harness: use harness.AgentConfig for preconfigured memory, skills, Todo, modes, approvals, history, and tracing with the normal provider constructors. Includes a Foundry example.
• Workflows: multi-agent patterns, routing, checkpointing, observability, and workflow orchestration
• End-to-End: full applications and demos
• Feature Comparison: .NET and Go SDK feature parity notes

Community & Feedback

• Found a bug? File a GitHub issue to help us improve.
• Enjoying MAF for Go? to show your support and help others discover the project.
• Have questions? Join our Discord.

Troubleshooting

Authentication

Tip: DefaultAzureCredential is convenient for development but in production, consider using a specific credential, such as managed identity, to avoid latency issues, unintended credential probing, and potential security risks from fallback mechanisms.

Environment Variables

For environment variable configuration specific to each sample, refer to the README or source in the sample directory under examples.

Contributor Resources

• Contributing Guide
• Code of Conduct
• Security Policy
• Support Policy
• Responsible AI Transparency FAQ

Important Notes

[!IMPORTANT] If you use Microsoft Agent Framework to build applications that operate with any third-party servers, agents, code, or non-Azure Direct models ("Third-Party Systems"), you do so at your own risk. Third-Party Systems are Non-Microsoft Products under the Microsoft Product Terms and are governed by their own third-party license terms. You are responsible for any usage and associated costs.

We recommend reviewing all data being shared with and received from Third-Party Systems and being cognizant of third-party practices for handling, sharing, retention, and location of data. It is your responsibility to manage whether your data will flow outside of your organization's Azure compliance and geographic boundaries and any related implications, and that appropriate permissions, boundaries, and approvals are provisioned.

You are responsible for carefully reviewing and testing applications you build using Microsoft Agent Framework in the context of your specific use cases, and making all appropriate decisions and customizations. This includes implementing your own responsible AI mitigations such as metaprompt, content filters, or other safety systems, and ensuring your applications meet appropriate quality, reliability, security, and trustworthiness standards. See also: Transparency FAQ.

Telemetry and data collection

This repository does not configure a Microsoft telemetry exporter. The OpenAI/Foundry adapters include SDK identity/version in their User-Agent and append a process-wide feature-usage bitmask only on approved HTTPS Azure/Foundry destinations. The mask records which framework capabilities were exercised, not invocation counts, prompts, arguments, or user data. It is not sent to third-party destinations or added to OpenTelemetry.

Set AGENT_FRAMEWORK_FEATURE_MASK_DISABLED=true or 1 before first use to disable feature tracking and emission while retaining the base SDK User-Agent. Set AGENT_FRAMEWORK_USER_AGENT_DISABLED=true or 1 to suppress the framework's entire User-Agent contribution, including the feature mask. See the feature-usage registry and emission policy.

Some packages and samples include optional OpenTelemetry instrumentation or connect to Microsoft-hosted services. To turn off that instrumentation, do not configure OpenTelemetry exporters or telemetry middleware in your application. Service telemetry, if any, is governed by the services you choose to call.

Preview status

The Agent Framework for Go is in public preview. Declarative agents, RAG, CodeAct, and functional workflows are not yet available. File issues on GitHub at https://github.com/microsoft/agent-framework-go/issues.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos is subject to those third-party policies.