# Microsoft Agent Framework vs Google ADK

> Source: https://theagentsindex.com/compare/google-adk-vs-microsoft-agent-framework (structured, researched, re-verified)

Microsoft Agent Framework (MAF) and Google Agent Development Kit (ADK) answer the same question from two different hyperscalers: should a major AI lab ship its own official, opinionated agent SDK instead of leaving the category to third parties like LangChain or CrewAI? Both say yes, but they arrived from opposite directions. MAF is explicitly a consolidation. Microsoft merged two previously separate stacks, AutoGen's multi-agent orchestration and Semantic Kernel's enterprise plumbing, into one supported framework, and put both predecessors into maintenance mode (bug and security fixes only) to push new development toward it. It reached a production-ready 1.0 GA on 2026-04-03. ADK is a from-scratch product: Google announced it at Cloud NEXT in April 2025, a full year earlier, and built it around protocol interoperability and evaluation tooling from day one rather than merging legacy stacks.

Checked live via the GitHub API on 2026-08-06: Microsoft's repository (microsoft/agent-framework) carries 12,622 stars and 2,117 forks; Google's flagship repository (google/adk-python) carries 21,013 stars and 3,803 forks. Both were pushed to within the last day of this research, real active projects rather than abandoned ones. Language spread is the clearest structural difference between them. MAF ships production-grade SDKs in three languages (.NET, Python and Go), reflecting its .NET/Azure-enterprise roots; ADK ships four (Python, Go, Java and TypeScript, all reaching 1.0 GA) plus an experimental Kotlin/Android runtime, the broadest spread of any framework in this index. A team standardizing on Java or TypeScript has a first-party GA SDK on ADK and none on MAF.

Both ship native support for the two dominant 2026 agent-interoperability standards, the Agent2Agent (A2A) protocol and the Model Context Protocol (MCP), so agents built on either framework can in principle hand work to agents built on the other. Where they diverge is which production concern each treats as first-class. MAF's Agent Harness gives an agent guarded shell and filesystem access, human-in-the-loop approval flows and automatic context compaction; its CodeAct mode, where the model writes a short program to batch tool calls inside an isolated micro-VM instead of calling them one at a time, is the one benchmarked performance claim in this comparison, with Microsoft's own BUILD 2026 numbers reporting 52.4% faster completion and 63.9% fewer tokens on multi-step workloads. ADK's built-in strength is evaluation: `adk eval` scores an agent's full trajectory and final output against repeatable test cases, and a local ADK Web UI debugs a flow before shipping it, a stronger built-in testing story than MAF ships today, though ADK has no equivalent to the Agent Harness's guarded real-environment access.

Pricing is free-versus-free at the open-source core: MAF is MIT-licensed, ADK is Apache 2.0, and both are fully self-hostable at no cost beyond the model provider's own bill. The managed layers price differently. MAF's Foundry Hosted Agents publishes no flat fee at all; it bills purely as ordinary Azure/Microsoft Foundry consumption, so there is no separate number to compare. ADK's Agent Runtime publishes explicit per-unit metering instead: $0.0864 per vCPU-hour plus $0.0090 per GB-hour, with a free allowance of 50 vCPU-hours and 100 GB-hours per month, plus $0.25 per 1,000 events for session and memory storage. Neither managed layer is required to use the framework; both cores run anywhere for free.

| Fact | Microsoft Agent Framework | Google Agent Development Kit |
| --- | --- | --- |
| What it is | Microsoft's unified, MIT-licensed agent SDK for .NET and Python: the supported successor to AutoGen and Semantic Kernel. | Google's open-source, code-first agent SDK with five language runtimes (Python, TypeScript, Go, Java, Kotlin), native A2A and MCP support, and an optional managed runtime on Google Cloud. |
| Pricing | Free | Free |
| API | Yes — The framework exposes a public SDK API for building agents from your own code: the Python package `agent-framework` on PyPI (pip install agent-framework, core and per-integration sub-packages) and the .NET package `Microsoft.Agents.AI` on NuGet (dotnet add package Microsoft.Agents.AI, plus Microsoft.Agents.AI.Foundry for Foundry integration). A Go SDK ships as github.com/microsoft/agent-framework-go. No public REST API of its own; agents built with the SDK consume LLM provider APIs directly. | Yes |
| Tags | API · Claude · Free tier · GPT · Multi-agent · Open models · Open source · Self-hosted | API · Claude · Free tier · GPT · Multi-agent · Open models · Open source · Self-hosted |
| Best for | .NET or Python teams already invested in Azure or Microsoft Foundry who want one supported agent SDK, an optional managed hosting path, and a documented migration off AutoGen or Semantic Kernel. | Teams that want Google's own official, actively-developed agent SDK, especially polyglot teams (Python/Go/Java/TypeScript/Kotlin) or anyone building multi-agent systems that need to interoperate across vendors via the A2A protocol. |
| Full facts | https://theagentsindex.com/microsoft-agent-framework.md | https://theagentsindex.com/google-adk.md |

## Verdict: Microsoft Agent Framework or Google Agent Development Kit?

Already standardized on Azure or Microsoft Foundry, or migrating an existing AutoGen or Semantic Kernel codebase, pick Microsoft Agent Framework: it has a documented first-party migration path and is the supported successor rather than an ambiguous third option. Already on Google Cloud or Gemini, or building a genuinely polyglot system across Python, Go, Java and TypeScript, pick Google ADK: it is the only framework in this category shipping four production-grade language SDKs. Need an agent that can safely act inside a real environment today, with shell access, filesystem access and human approval, backed by a measured performance benchmark: MAF's Agent Harness and CodeAct are further along than anything ADK ships natively. Need a strong built-in evaluation loop before trusting an agent in production: ADK's `adk eval` and ADK Web are purpose-built for exactly that, ahead of MAF's current tooling. Neither commits a team to its parent cloud. MAF reaches Anthropic Claude and local Ollama models alongside Azure OpenAI; ADK reaches OpenAI, Claude, Cohere and local models via LiteLLM alongside Gemini. The honest deciding factor is which hyperscaler ecosystem, and which language line-up, a team is already committed to.

## FAQ

**Are Microsoft Agent Framework and Google ADK direct competitors?**

Yes — both listings on this site already name the other as their closest comparison: Microsoft Agent Framework's alternatives rail describes Google ADK as “the closest analogue in this index,” and ADK's rail reciprocally names Microsoft Agent Framework first. Both are official, first-party agent SDKs from a major AI lab, built to compete with third-party frameworks like LangChain and CrewAI as much as with each other.

**Which has more GitHub traction?**

Google ADK leads, checked live via the GitHub API on 2026-08-06: 21,013 stars and 3,803 forks on google/adk-python, versus 12,622 stars and 2,117 forks on microsoft/agent-framework. Context cuts the other way on age, though: ADK was announced a full year before MAF reached its 1.0 GA (April 2025 vs April 2026), and MAF's own predecessor, AutoGen, separately carries roughly 59.8k stars on its now-frozen repository — a larger legacy codebase Microsoft is deliberately moving developers away from.

**Are both actually free?**

Yes, at the core. Microsoft Agent Framework is MIT-licensed; Google ADK is Apache 2.0. Both are fully self-hostable anywhere for free, paying only the model provider. The managed layers differ: MAF's Foundry Hosted Agents has no published flat fee, billed purely as Azure/Microsoft Foundry consumption; ADK's Agent Runtime publishes explicit metering ($0.0864/vCPU-hour + $0.0090/GB-hour, with a free monthly allowance).

**Which programming languages does each support?**

Microsoft Agent Framework ships production-grade 1.0 GA SDKs for .NET, Python and Go. Google ADK ships four: Python, Go, Java and TypeScript, all at 1.0 GA, plus an experimental Kotlin/Android runtime — the broadest language spread of any framework in this index. A team standardizing on Java or TypeScript has no first-party GA option on MAF today.

**Do either lock a team into their parent cloud or model?**

No, both are genuinely multi-provider despite the branding. Microsoft Agent Framework reaches Azure OpenAI, OpenAI, Anthropic Claude and local Ollama models as first-class providers, plus Google Gemini via its Go SDK. Google ADK is optimized for Gemini but reaches OpenAI, Anthropic Claude, Cohere and 100+ other providers through a LiteLLM integration. Each framework's managed hosting layer (Foundry Hosted Agents / Agent Runtime) is optional, not required to run either SDK.