Agent frameworks - verified
AI agent frameworks compared
16 frameworks for building AI agents, side by side: the languages each one ships for, its license, the latest published version, and whether its own docs document MCP, multi-agent orchestration, durable execution, human in the loop, streaming, structured output, tracing, memory, evals and swapping model providers. 14 support Python and 8 TypeScript. Install commands are copied verbatim from official docs. Want our opinionated picks instead? See the best AI frameworks.
- Frameworks
- 16
- Python
- 14
- TypeScript
- 8
- Document MCP
- 15
Feature matrix
Yes and No link to the doc that says so. Not stated means the docs do not say either way. Scroll the table sideways on a small screen.
16 frameworks
Framework by framework
What each framework is, as its own docs describe it. Each page has the install commands, a quickstart, the feature table with a source for every value, and the comparisons it appears in.
LangGraph
1.2.16Python, TypeScript
LangGraph is a low-level orchestration framework and runtime from LangChain Inc for building, managing, and deploying long-running, stateful agents, and it can be used with or without LangChain.
LangChain (agents)
1.4.6Python, TypeScript
LangChain is a framework for building agents and LLM-powered applications whose create_agent function provides a configurable agent harness of model, tools, prompt, and middleware, built on top of LangGraph.
CrewAI
1.15.28Python
CrewAI is an open-source Python framework for building multi-agent workflows that combines Crews of role-based autonomous agents with event-driven Flows that manage state and control execution.
LlamaIndex (agent workflows)
0.14.25Python
LlamaIndex is an open-source framework for building agentic applications whose Agent Workflows provide an event-driven, step-based way to control execution, alongside prebuilt agents such as FunctionAgent and the multi-agent AgentWorkflow.
OpenAI Agents SDK
0.24.0Python, TypeScript
The OpenAI Agents SDK is an open-source Python and TypeScript framework from OpenAI for building multi-agent workflows from a small set of primitives: agents, agents as tools or handoffs, and guardrails, with built-in tracing.
Claude Agent SDK
0.2.168Python, TypeScript
The Claude Agent SDK is a Python and TypeScript library from Anthropic that gives developers the same tools, agent loop, and context management that power Claude Code, for running agents inside their own applications.
Python, TypeScript, Go, Java, Other
Agent Development Kit (ADK) is Google's open-source, code-first framework for building, evaluating, and deploying AI agents, available in Python, TypeScript, Go, Java, and Kotlin, optimized for Gemini and described by its docs as model-agnostic.
Python, C#, Go
Microsoft Agent Framework is an open-source framework from Microsoft for building AI agents and graph-based multi-agent workflows in .NET and Python, with a Go SDK in public preview, and is the direct successor to Semantic Kernel and AutoGen.
Vercel AI SDK (agents)
7.0.143TypeScript
The AI SDK is Vercel's provider-agnostic TypeScript toolkit for building AI applications and agents, and its ToolLoopAgent class runs a language model with tools in a loop until a task is complete.
Mastra
1.78.0TypeScript
Mastra is a TypeScript framework for building AI agents and applications, combining agents, graph-based workflows, memory, evals and observability, with models addressed through a provider/model router.
Pydantic AI
2.57.0Python
Pydantic AI is a Python agent framework from the Pydantic team that provides a typed agent loop with structured outputs, typed tools and dependency injection, and lets the model be swapped across many providers with a string.
Python
smolagents is an open-source Python library from Hugging Face for building and running agents in a few lines of code, with first-class support for CodeAgent, which writes its actions as Python code, alongside a conventional ToolCallingAgent.
Strands Agents
1.61.1Python, TypeScript
Strands Agents is an open-source SDK for building and running AI agents in Python and TypeScript inside your own process, with no hosted control plane, plus an optional preassembled agent harness.
Agno
3.1.4Python
Agno is a Python framework and runtime for agent platforms: the Agno SDK builds agents, teams and workflows, the AgentOS runtime serves them as an API, and the AgentOS UI manages them.
DSPy
3.4.1Python
DSPy is a Python framework for building AI systems by declaring tasks as typed signatures composed into modules, with optimizers that tune prompts and weights against a metric.
Python
AG2 is an open-source, async Python framework for building AI agents and multi-agent networks that diverged from the original AutoGen codebase in November 2024; since v1.0 the classic autogen API lives in the separate AG2 Classic package.
Compare two frameworks
Side by side, field by field, with the differences marked and both install commands.
- OpenAI Agents SDK vs Google Agent Development Kit (ADK)
- CrewAI vs AG2 (formerly AutoGen)
- LangGraph vs Google Agent Development Kit (ADK)
- Pydantic AI vs LangGraph
- Mastra vs Vercel AI SDK (agents)
- Strands Agents vs OpenAI Agents SDK
- Agno vs CrewAI
- smolagents (Hugging Face) vs Pydantic AI
- Microsoft Agent Framework vs LangGraph
- LlamaIndex (agent workflows) vs LangGraph
- DSPy vs LangChain (agents)
- Claude Agent SDK vs Google Agent Development Kit (ADK)
Already compared elsewhere on the site
- LangGraph vs CrewAITool comparison
- LangChain vs LlamaIndexTool comparison
- Claude Agent SDK vs LangGraphArticle
- Mastra vs LangGraph.jsArticle
- OpenAI Agents SDK vs Claude Agent SDKArticle
- Claude Agent SDK vs OpenAI Agents SDKTool comparison
- Vercel AI SDK vs LangGraphArticle
- Vercel AI SDK vs LangChainTool comparison
- LangChain vs OpenAI Agents SDKTool comparison
- Vercel AI SDK vs Claude Agent SDKTool comparison
Method
How to read this table
Read from each framework's official docs, README, package registry and release pages on . Frameworks ship weekly; check the linked source before you rely on a feature.
- Yes, No, Not stated
- Yes and No appear only when the framework's own docs say so, and each links to the page that says it. Not stated means the docs neither claim nor rule out the feature; it is not a No.
- Commands copied verbatim
- Install commands and quickstart snippets come from the official page linked under each one, unchanged. Versions come from PyPI, npm, the Go module proxy, Maven Central, NuGet or GitHub releases.
- Not a quality ranking
- Nothing here measures speed, reliability or developer experience. The table lists documented capabilities only; the order follows the data file, not a score.
FAQ
Which agent frameworks support MCP?
15 of 16 document Model Context Protocol (MCP) support: LangChain (agents), CrewAI, LlamaIndex (agent workflows), OpenAI Agents SDK, Claude Agent SDK, Google Agent Development Kit (ADK), Microsoft Agent Framework, Vercel AI SDK (agents), Mastra, Pydantic AI, smolagents (Hugging Face), Strands Agents, Agno, DSPy and AG2 (formerly AutoGen). The docs of LangGraph do not say. Each Yes links to the page that says so.
Which agent frameworks support durable execution?
14 of 16 document durable execution (runs that persist state and resume after a failure or pause): LangGraph, LangChain (agents), CrewAI, LlamaIndex (agent workflows), OpenAI Agents SDK, Claude Agent SDK, Google Agent Development Kit (ADK), Microsoft Agent Framework, Vercel AI SDK (agents), Mastra, Pydantic AI, Strands Agents, Agno and AG2 (formerly AutoGen). The docs of smolagents (Hugging Face) and DSPy do not say. Each Yes links to the page that says so.
Which agent frameworks support human-in-the-loop approval?
15 of 16 document human-in-the-loop support: LangGraph, LangChain (agents), CrewAI, LlamaIndex (agent workflows), OpenAI Agents SDK, Claude Agent SDK, Google Agent Development Kit (ADK), Microsoft Agent Framework, Vercel AI SDK (agents), Mastra, Pydantic AI, smolagents (Hugging Face), Strands Agents, Agno and AG2 (formerly AutoGen). The docs of DSPy do not say. Each Yes links to the page that says so.
Which agent frameworks support multi-agent systems?
15 of 16 document multi-agent orchestration: LangGraph, LangChain (agents), CrewAI, LlamaIndex (agent workflows), OpenAI Agents SDK, Claude Agent SDK, Google Agent Development Kit (ADK), Microsoft Agent Framework, Vercel AI SDK (agents), Mastra, Pydantic AI, smolagents (Hugging Face), Strands Agents, Agno and AG2 (formerly AutoGen). The docs of DSPy do not say. Each Yes links to the page that says so.
Which agent frameworks support Python and which support TypeScript?
14 of 16 list Python and 8 list TypeScript. Both: LangGraph, LangChain (agents), OpenAI Agents SDK, Claude Agent SDK, Google Agent Development Kit (ADK) and Strands Agents. TypeScript only: Vercel AI SDK (agents) and Mastra. Python without TypeScript: CrewAI, LlamaIndex (agent workflows), Microsoft Agent Framework, Pydantic AI, smolagents (Hugging Face), Agno, DSPy and AG2 (formerly AutoGen). Other languages appear on each framework's page.
Which agent frameworks are model-agnostic?
14 of 16 document working with models from more than one provider: LangChain (agents), CrewAI, LlamaIndex (agent workflows), OpenAI Agents SDK, Google Agent Development Kit (ADK), Microsoft Agent Framework, Vercel AI SDK (agents), Mastra, Pydantic AI, smolagents (Hugging Face), Strands Agents, Agno, DSPy and AG2 (formerly AutoGen). The docs of LangGraph and Claude Agent SDK do not say. Each Yes links to the page that says so.
Which agent frameworks have built-in evals?
10 of 16 document evaluation (evals) support: CrewAI, LlamaIndex (agent workflows), Google Agent Development Kit (ADK), Microsoft Agent Framework, Mastra, Pydantic AI, Strands Agents, Agno, DSPy and AG2 (formerly AutoGen). The docs of LangGraph, LangChain (agents), OpenAI Agents SDK, Claude Agent SDK, Vercel AI SDK (agents) and smolagents (Hugging Face) do not say. Each Yes links to the page that says so.
Which licenses do these agent frameworks use?
MIT: LangGraph, LangChain (agents), CrewAI, LlamaIndex (agent workflows), OpenAI Agents SDK, Microsoft Agent Framework, Pydantic AI and DSPy. Apache-2.0: Google Agent Development Kit (ADK), Vercel AI SDK (agents), smolagents (Hugging Face), Strands Agents, Agno and AG2 (formerly AutoGen). Anthropic Commercial Terms of Service (the Python SDK repository is MIT licensed): Claude Agent SDK. Apache-2.0 (code in ee/ directories is under the Mastra Enterprise License): Mastra. Each framework's page links its license.
What does "Not stated" mean in the table?
That the framework's official docs, read on October 18, 2026, neither claim nor rule out the feature. It is not a No: the feature may exist without being documented. Yes and No are shown only when the docs say so, and each one links to the page that says it.
Is this a ranking of the best agent framework?
No. The table lists what each framework's docs say it supports, how to install it and its latest version. It measures neither speed nor quality, and the order follows the data file, not a score.
Newsletter
Agent frameworks, every week
One email per week. Framework releases, new agent tooling and what they mean for what you build.