--- slug: "agent-frameworks-compared" title: "AI agent frameworks compared" description: "LangGraph, CrewAI, OpenAI Agents SDK, Pydantic AI, Google ADK, Microsoft Agent Framework, Mastra and more by language, control model, owner and license." published: "2026-10-15" updated: "2026-10-15" signals_fetched_at: "2026-10-15T05:02:49.930Z" tools: ["mastra", "google-adk", "langgraph", "microsoft-agent-framework", "agno", "claude-agent-sdk", "crewai", "deep-agents", "openai-agents-sdk", "pydantic-ai", "smolagents", "strands-agents", "langchain", "llamaindex", "vercel-ai-sdk", "autogen"] page: "https://agenttooldrop.com/guides/agent-frameworks-compared" --- # AI agent frameworks compared Updated 2026-10-15. Page: https://agenttooldrop.com/guides/agent-frameworks-compared The frameworks and SDKs in our catalog that you write agent code against, grouped by what they optimize for: explicit control over flow and state, a fast start with opinionated pieces, a small typed or minimal loop, or a large integration ecosystem. The table gives the language, the control model in a phrase, the GitHub owner and the license. Every description restates the project's own README or docs. Nothing here is a benchmark. Language usually narrows the field first; after that, the question is how much of the control flow you want to own. ## At a glance | Tool | Language | Control model | GitHub owner | License | Stars | Last push | | --- | --- | --- | --- | --- | --- | --- | | [Mastra](https://agenttooldrop.com/tools/mastra) | TypeScript | Agents plus graph workflows with suspend and resume | mastra-ai | Apache-2.0 | 28,827 | 2026-10-15 | | [Google Agent Development Kit (ADK)](https://agenttooldrop.com/tools/google-adk) | Python 3.11+; Java, Kotlin, Go and TypeScript in other repos | Agent hierarchies and graph workflows, with eval and deploy commands | google | Apache-2.0 | 21,854 | 2026-10-15 | | [LangGraph](https://agenttooldrop.com/tools/langgraph) | Python; LangGraph.js for JS/TS | Low-level graphs with durable execution and human-in-the-loop | langchain-ai | MIT | 43,287 | 2026-10-15 | | [Microsoft Agent Framework](https://agenttooldrop.com/tools/microsoft-agent-framework) | Python, .NET; Go in preview in a separate repo | Graph workflows: sequential, concurrent, handoff, group; checkpointing | microsoft | MIT | 14,132 | 2026-10-14 | | [Agno](https://agenttooldrop.com/tools/agno) | Python | Agents, teams and workflows served by the AgentOS runtime | agno-agi | Apache-2.0 | 42,736 | 2026-10-15 | | [Claude Agent SDK](https://agenttooldrop.com/tools/claude-agent-sdk) | Python, TypeScript | Claude Code's agent loop and built-in tools as a library | anthropics | Python repo MIT | 8,283 | 2026-10-14 | | [CrewAI](https://agenttooldrop.com/tools/crewai) | Python 3.10 to 3.13 | Role-based crews plus event-driven flows | crewAIInc | MIT | 59,704 | 2026-10-15 | | [Deep Agents](https://agenttooldrop.com/tools/deep-agents) | Python; deepagentsjs for JS/TS | Ready-made harness on LangGraph: planning, files, subagents | langchain-ai | MIT | 30,277 | 2026-10-15 | | [OpenAI Agents SDK](https://agenttooldrop.com/tools/openai-agents-sdk) | Python 3.10+; a separate JS/TS SDK | Agents with tools, handoffs and guardrails; built-in tracing | openai | MIT | 30,042 | 2026-10-15 | | [Pydantic AI](https://agenttooldrop.com/tools/pydantic-ai) | Python | Typed agent loop with validated tools and outputs | pydantic | MIT | 20,622 | 2026-10-15 | | [smolagents](https://agenttooldrop.com/tools/smolagents) | Python | Agents that write their actions as Python code | huggingface | Apache-2.0 | 29,849 | 2026-10-06 | | [Strands Agents](https://agenttooldrop.com/tools/strands-agents) | Python 3.10+, TypeScript (Node.js 22+) | Model-driven loop in your process, with hooks at each step | strands-agents | Apache-2.0 | 8,817 | 2026-10-14 | | [LangChain](https://agenttooldrop.com/tools/langchain) | Python; LangChain.js for JS/TS | Component interfaces and integrations; LangGraph for orchestration | langchain-ai | MIT | 147,843 | 2026-10-14 | | [LlamaIndex](https://agenttooldrop.com/tools/llamaindex) | Python | Data connectors, indexes and query engines, plus event-driven workflows | run-llama | MIT | 52,501 | 2026-10-14 | | [Vercel AI SDK](https://agenttooldrop.com/tools/vercel-ai-sdk) | TypeScript (Node.js 22+) | One API for text, structured output and tool-calling agents, plus UI hooks | vercel | Apache-2.0 | 27,323 | 2026-10-15 | | [AutoGen](https://agenttooldrop.com/tools/autogen) | Python | Event-driven multi-agent apps; no new features | microsoft | MIT for code | 61,411 | 2026-04-15 | ## Explicit control over flow and state ### Mastra Builds TypeScript agents and graph-based workflows with memory, RAG, evals and model routing Pick it if you want agents, workflows, memory and evals in one TypeScript framework. License: Apache-2.0; code in ee/ directories under the Mastra Enterprise License (source-available). Pricing: Open source. GitHub: 28,827 stars, last push 2026-10-15. Skip it if: You work in Python. Features in ee/ directories (for example auth) need a Mastra enterprise license for production use. Create a project with npm: ```sh npm create mastra@latest ``` Install source: https://github.com/mastra-ai/mastra#readme. Entry: https://agenttooldrop.com/tools/mastra.md ### Google Agent Development Kit (ADK) Builds, evaluates and deploys Python agents and graph-based workflows, optimized for Gemini Pick it if you want agent definitions, a local testing UI, evaluation and deployment to Cloud Run in one toolkit. License: Apache-2.0. Pricing: Open source. GitHub: 21,854 stars, last push 2026-10-15. Skip it if: You are on Python older than 3.11. If Python is not your stack, ADK has separate Java, Kotlin, Go and TypeScript versions in other repos. Install with pip: ```sh pip install google-adk ``` Install source: https://github.com/google/adk-python#readme. Entry: https://agenttooldrop.com/tools/google-adk.md ### LangGraph Orchestrates long-running, stateful agents as graphs with durable execution and human-in-the-loop Pick it if you want explicit control over control flow and state, with checkpoints to resume from after a failure or a human review. License: MIT. Pricing: Open source. GitHub: 43,287 stars, last push 2026-10-15. Skip it if: You want a quick agent with planning, subagents and file system use out of the box; the README points to Deep Agents, a higher-level package built on LangGraph, for that. Install with pip: ```sh pip install -U langgraph ``` Install source: https://github.com/langchain-ai/langgraph#readme. Entry: https://agenttooldrop.com/tools/langgraph.md ### Microsoft Agent Framework Builds AI agents and graph-based multi-agent workflows in Python or .NET Pick it if you need multi-agent workflows with checkpointing and human review in Python or C#, or you are moving off AutoGen or Semantic Kernel. License: MIT. Pricing: Open source. GitHub: 14,132 stars, last push 2026-10-14. Skip it if: You want a minimal single-agent loop with no Azure ties; the README quickstarts use Microsoft Foundry and Azure credentials, and other providers are covered in samples. The Go SDK is a separate repo in public preview. Install with pip: ```sh pip install agent-framework ``` Install source: https://github.com/microsoft/agent-framework#readme. Entry: https://agenttooldrop.com/tools/microsoft-agent-framework.md ## A fast start with opinionated pieces ### Agno Builds Python agents, teams and workflows and serves them as an API with the AgentOS runtime Pick it if you want to run your agents as a multi-user service on your own infrastructure, with data in your own database. License: Apache-2.0. Pricing: Open source. GitHub: 42,736 stars, last push 2026-10-15. Skip it if: You are not on Python. Agno sends a telemetry event per agent run by default (prompts, messages and outputs are not sent); set AGNO_TELEMETRY=false to turn it off. Install with pip: ```sh pip install -U agno openai ``` Install source: https://docs.agno.com/sdk/setup. Entry: https://agenttooldrop.com/tools/agno.md ### Claude Agent SDK Embeds Claude Code's agent loop, built-in tools and context management in Python or TypeScript apps Pick it if you want Claude Code's tools and permission model inside your own process. It runs Claude models only. License: Python repo MIT; use governed by Anthropic Commercial Terms of Service. Pricing: Paid. GitHub: 8,283 stars, last push 2026-10-14. Skip it if: You want to use non-Claude models. It needs an Anthropic API key or Claude through Amazon Bedrock, Google Cloud or Microsoft Foundry, and third-party products may not offer claude.ai login unless approved by Anthropic. Install with pip: ```sh pip install claude-agent-sdk ``` Install source: https://github.com/anthropics/claude-agent-sdk-python#readme. Entry: https://agenttooldrop.com/tools/claude-agent-sdk.md ### CrewAI Orchestrates teams of role-based agents (Crews) and event-driven workflows (Flows) in Python Pick it if you want specialized agents with clear roles and tasks, wrapped in event-driven workflows in Python. License: MIT. Pricing: Open source. GitHub: 59,704 stars, last push 2026-10-15. Skip it if: You are not on Python (it requires Python 3.10 to 3.13). It collects anonymous telemetry by default (disable with OTEL_SDK_DISABLED=true), and the managed control plane, CrewAI AMP, is a separate commercial product. Install the CLI with uv: ```sh uv tool install crewai ``` Install source: https://github.com/crewAIInc/crewAI#readme. Entry: https://agenttooldrop.com/tools/crewai.md ### Deep Agents Gives you a ready-made agent harness with planning, files, subagents and context management Pick it if you want planning, subagents, files and context management on day one, with every piece overridable. License: MIT. Pricing: Open source. GitHub: 30,277 stars, last push 2026-10-15. Skip it if: You need a custom control flow; LangGraph underneath gives you that directly. Its security model trusts the LLM, so you must enforce limits at the tool or sandbox level. Install with uv: ```sh uv add deepagents ``` Install source: https://github.com/langchain-ai/deepagents#readme. Entry: https://agenttooldrop.com/tools/deep-agents.md ### OpenAI Agents SDK Builds multi-agent workflows with handoffs, guardrails, sessions and built-in tracing Pick it if you want a small set of primitives (agents, tools, handoffs, guardrails) with tracing built in. License: MIT. Pricing: Open source. GitHub: 30,042 stars, last push 2026-10-15. Skip it if: You work in JavaScript or TypeScript; use the separate Agents SDK JS/TS (openai/openai-agents-js) instead. The Python package needs Python 3.10 or newer. Install with pip: ```sh pip install openai-agents ``` Install source: https://github.com/openai/openai-agents-python#readme. Entry: https://agenttooldrop.com/tools/openai-agents-sdk.md ## A small loop, typed or minimal ### Pydantic AI Runs a typed Python agent loop with validated tools and outputs, and models swapped by a string Pick it if you want type checking and validation across tools and outputs, and to switch providers without rewriting the agent. License: MIT. Pricing: Open source. GitHub: 20,622 stars, last push 2026-10-15. Skip it if: You are not working in Python. Install with uv: ```sh uv add pydantic-ai ``` Install source: https://github.com/pydantic/pydantic-ai#readme. Entry: https://agenttooldrop.com/tools/pydantic-ai.md ### smolagents Runs agents that write their actions as Python code, with core agent logic in about 1,000 lines Pick it if you want a small, readable library whose agents act by writing code, and you can run that code in a sandbox. License: Apache-2.0. Pricing: Open source. GitHub: 29,849 stars, last push 2026-10-06. Skip it if: You cannot run agent code in a sandbox: the README says the built-in LocalPythonExecutor is not a security boundary and recommends E2B, Blaxel, Modal or Docker. Install with pip: ```sh pip install "smolagents[toolkit]" ``` Install source: https://github.com/huggingface/smolagents#readme. Entry: https://agenttooldrop.com/tools/smolagents.md ### Strands Agents Runs a model-driven agent loop in your own process, in Python or TypeScript, with any model provider Pick it if you want to own the agent loop in your process, with limits, hooks and tracing built in. License: Apache-2.0. Pricing: Open source. GitHub: 8,817 stars, last push 2026-10-14. Skip it if: You want a hosted agent platform rather than a library. The Python SDK needs Python 3.10+ and the TypeScript SDK needs Node.js 22+. Install with pip: ```sh pip install strands-agents strands-agents-tools ``` Install source: https://github.com/strands-agents/harness-sdk#readme. Entry: https://agenttooldrop.com/tools/strands-agents.md ## Integrations, data and web apps ### LangChain Provides one interface and a large integration library for models, embeddings, vector stores and tools Pick it if you want ready-made integrations for models, vector stores and retrievers behind one interface. License: MIT. Pricing: Open source. GitHub: 147,843 stars, last push 2026-10-14. Skip it if: You need fine-grained control over agent flow and state; the README points to LangGraph for advanced customization and agent orchestration. Install with pip: ```sh pip install -U langchain ``` Install source: https://docs.langchain.com/oss/python/langchain/install. Entry: https://agenttooldrop.com/tools/langchain.md ### LlamaIndex Connects LLM agents to your data with connectors, indexes, query engines and event-driven workflows Pick it if your agent's main job is answering questions or taking actions over your own documents and data sources. License: MIT. Pricing: Open source. GitHub: 52,501 stars, last push 2026-10-14. Skip it if: You want a framework that is its maintainer's main focus: the README says the company's primary focus has shifted to LlamaParse, while the OSS framework stays available. Install with pip: ```sh pip install llama-index ``` Install source: https://developers.llamaindex.ai/python/framework/getting_started/installation/. Entry: https://agenttooldrop.com/tools/llamaindex.md ### Vercel AI SDK Gives TypeScript apps one API for text, structured output, tool-calling agents and chat UI hooks Pick it if you are building an agent into a web app and want the model calls and the chat UI from one library. License: Apache-2.0. Pricing: Open source. GitHub: 27,323 stars, last push 2026-10-15. Skip it if: You are not on JavaScript or TypeScript; it needs Node.js 22+. If you do not want the default Vercel AI Gateway routing, plan to install provider packages such as @ai-sdk/openai. Install with npm: ```sh npm install ai ``` Install source: https://github.com/vercel/ai#readme. Entry: https://agenttooldrop.com/tools/vercel-ai-sdk.md ## In maintenance mode ### AutoGen Builds multi-agent AI apps that act autonomously or alongside humans; now in maintenance mode Only if you already run it: AutoGen is in maintenance mode, and Microsoft recommends Microsoft Agent Framework for new projects. License: MIT for code (LICENSE-CODE); CC-BY-4.0 for docs and other content. Pricing: Open source. GitHub: 61,411 stars, last push 2026-04-15. Skip it if: You are starting a new project: AutoGen is in maintenance mode and community managed, and Microsoft recommends its successor, Microsoft Agent Framework, which has a migration guide from AutoGen. Install with pip: ```sh pip install -U "autogen-agentchat" "autogen-ext[openai]" ``` Install source: https://github.com/microsoft/autogen#readme. Entry: https://agenttooldrop.com/tools/autogen.md ## Newer options worth a look New and rising tools from the catalog that this guide does not compare yet, by new and rising score. Younger and less proven than the picks above. - [Reef](https://agenttooldrop.com/tools/reef.md): Turns feedback on an agent's live requests into versioned updates to its model weights or its harness. Skip it if: You want a stable agent today: harness work needs a model endpoint, representative tasks and an evaluator, weight training needs a supported GPU stack, and it is at v0.1.3 with the repo created on 2026-08-31. The harness install pipes a script into bash and the pi adapter runs commands with no sandbox, so keep the install root outside the project the agent works in. - [OpenBot](https://agenttooldrop.com/tools/openbot.md): Runs self-hosted AI coworkers with their own browser, files and shell, each action policy-checked and logged. Skip it if: You want a finished product: the README calls it an alpha template to clone and change, with no hosted version. It needs Docker, Bun, a model key and a CopilotKit Intelligence project (a free plan exists, or you can self-host it), and Bots drive a browser with saved logins and run shell commands, so set up sign-in before anyone else can reach it. ## How we chose Picked from the 95 tools in our catalog, each verified against its own repository and docs; 16 made this guide. Within each group, new and rising tools come first, sorted by the new and rising score (star velocity, recent pushes and releases, repo age and an editorial weight), then the standards, the well-known tools, by name. It is not a benchmark. The number beside a tool is its rank in the full new and rising ledger, not within this guide; standards are not ranked. Install commands are copied from each project's docs; every tool page links to the page they came from. GitHub signals (stars, last push, latest release) were fetched on 2026-10-15. n8n is in the same catalog category but left out here because it is a visual workflow platform, not a library you write agents against. AutoGen is listed last because it no longer receives new features. The GitHub owner column is the organization that owns the repository we track.