Agent Frameworks

A curated list of popular AI agent frameworks and SDKs, grouped by ecosystem. Every entry links to its GitHub repository and shows a live star badge.

Governance capability legend

Four columns flag built-in Responsible-AI / security capabilities. This is a best-effort snapshot — features move fast, so verify against each project's current docs.

• Guardrails — native input/output validation, content/safety filters, policy checks, structured-output validation, or human-in-the-loop gates.
• Identity — a first-class agent identity or principal (stable agent ID, credentials, auth, RBAC) beyond a display name/role.
• Audit — built-in tracing/observability of agent steps (events, callbacks, OpenTelemetry).
• Sandbox — isolated execution of agent-generated code or tools (containers, microVMs, or a remote sandbox).

Cell values: ✅ built-in / first-class · 🟡 partial, basic, or via an official plugin/integration · — not built-in (still achievable with external components). Click a ✅ or 🟡 rating to open the supporting GitHub code or documentation.

Contents

• General-purpose & multi-agent orchestration
• Autonomous "AutoGPT-style" agents
• Low-code & visual builders
• JavaScript / TypeScript
• JVM (Java / Kotlin)
• Go
• Rust
• Coding agents
• Browser & computer use
• Voice & real-time
• Agent infrastructure (tools · memory · sandboxes)

---

General-purpose & multi-agent orchestration

Framework	Stars	Lang	Guardrails	Identity	Audit	Sandbox	What it is
LangChain		Python	🟡	—	✅	🟡	Composable framework for LLM apps with a huge tool/integration ecosystem
LangGraph		Python	🟡	—	✅	—	Stateful, graph-based multi-agent orchestration with human-in-the-loop
LlamaIndex		Python	🟡	—	✅	🟡	Data framework for RAG and agentic workflows
Microsoft AutoGen		Python/.NET	🟡	—	✅	✅	Multi-agent conversation framework
Semantic Kernel		C#/Py/Java	✅	🟡	✅	—	Enterprise agent SDK with plugins and invocation filters
Microsoft Agent Framework		.NET/Python	✅	✅	✅	🟡	Unified successor to Semantic Kernel + AutoGen (Entra Agent ID)
CrewAI		Python	✅	—	✅	🟡	Role-playing multi-agent "crews" and flows
OpenAI Agents SDK		Python	✅	—	✅	🟡	Lightweight agents with handoffs and guardrails
OpenAI Swarm		Python	—	—	—	—	Experimental/educational lightweight multi-agent orchestration
Google ADK		Python	✅	🟡	✅	✅	Google's model-agnostic Agent Development Kit
Pydantic AI		Python	🟡	—	✅	—	Type-safe agent framework from the Pydantic team
Haystack		Python	🟡	—	✅	—	Production LLM/RAG and agent pipelines (deepset)
Agno		Python	🟡	—	✅	—	High-performance multi-modal agents (formerly Phidata)
CAMEL		Python	—	—	🟡	🟡	Research framework for multi-agent societies
DSPy		Python	🟡	—	🟡	—	Program (don't prompt) LMs — optimizers and compilers
smolagents		Python	🟡	—	✅	✅	Minimal, code-first agents (Hugging Face)
Letta		Python	—	🟡	✅	—	Stateful agents with long-term memory (formerly MemGPT)
Langroid		Python	🟡	—	🟡	—	Multi-agent programming framework
Griptape		Python	🟡	—	✅	🟡	Modular agents, pipelines, and workflows
Marvin		Python	🟡	—	🟡	—	Lightweight AI engineering toolkit (Prefect)
Strands Agents		Python	🟡	🟡	✅	—	Model-driven agent SDK (AWS)
Atomic Agents		Python	🟡	—	🟡	—	Lightweight, composable "atomic" building blocks
AWS Agent Squad		Python/TS	—	—	🟡	—	Multi-agent orchestrator (formerly Multi-Agent Orchestrator)
BeeAI Framework		Python/TS	🟡	—	✅	🟡	Production agents (IBM · Linux Foundation)
Julep		Python	—	🟡	✅	✅	Serverless platform + SDK for stateful agents

Autonomous "AutoGPT-style" agents

Framework	Stars	Lang	Guardrails	Identity	Audit	Sandbox	What it is
AutoGPT		Python	🟡	—	🟡	✅	Pioneering autonomous GPT agent platform
MetaGPT		Python	🟡	—	🟡	—	Multi-agent "software company" framework
BabyAGI		Python	—	—	—	—	Minimal task-driven autonomous agent
SuperAGI		Python	—	—	🟡	—	Dev-first autonomous agent framework
AgentGPT		TypeScript	—	—	🟡	—	Browser-based autonomous agents

Low-code & visual builders

Framework	Stars	Lang	Guardrails	Identity	Audit	Sandbox	What it is
Dify		Python/TS	🟡	🟡	✅	✅	Open-source LLM app & agent platform
Flowise		TypeScript	🟡	🟡	🟡	🟡	Drag-and-drop LLM/agent builder
Langflow		Python	—	🟡	🟡	🟡	Visual builder for agent flows
n8n		TypeScript	—	🟡	✅	🟡	Workflow automation with native AI agent nodes
Rivet		TypeScript	—	—	🟡	—	Visual programming environment for AI agents
Rasa		Python	🟡	🟡	✅	—	Open-source conversational AI / assistants

JavaScript / TypeScript

Framework	Stars	Guardrails	Identity	Audit	Sandbox	What it is
LangChain.js		🟡	—	✅	🟡	JS/TS port of LangChain
Vercel AI SDK		🟡	—	✅	—	TS toolkit for AI apps/agents with UI streaming
Mastra		✅	—	✅	—	TS agent framework — workflows, memory, RAG
VoltAgent		✅	—	✅	—	TS agent framework with built-in observability
OpenAI Agents JS		✅	—	✅	🟡	Official JS/TS Agents SDK
Inngest AgentKit		—	—	🟡	—	Multi-agent networks on durable workflows

JVM (Java / Kotlin)

Framework	Stars	Guardrails	Identity	Audit	Sandbox	What it is
LangChain4j		✅	—	🟡	—	LangChain for Java
Spring AI		🟡	—	✅	—	Spring-native AI/agent framework
Google ADK (Java)		🟡	—	✅	✅	Agent Development Kit for the JVM

Go

Framework	Stars	Guardrails	Identity	Audit	Sandbox	What it is
Eino		🟡	—	✅	—	Go LLM/agent framework (CloudWeGo · ByteDance)
Genkit		🟡	—	✅	—	Cross-language agent framework (Go/JS/Python, Google)
LangChainGo		🟡	—	🟡	—	LangChain for Go

Rust

Framework	Stars	Guardrails	Identity	Audit	Sandbox	What it is
Rig		—	—	✅	—	Rust framework for LLM-powered agents

Coding agents

Framework	Stars	Guardrails	Identity	Audit	Sandbox	What it is
OpenHands		🟡	—	✅	✅	Autonomous software-engineering agents (formerly OpenDevin)
Aider		—	—	🟡	🟡	AI pair programming in your terminal
SWE-agent		🟡	—	✅	✅	Agents that resolve real GitHub issues (Princeton)
gpt-engineer		—	—	🟡	🟡	Build codebases from a natural-language prompt
Potpie		—	🟡	✅	—	Custom engineering agents for your codebase

Browser & computer use

Framework	Stars	Guardrails	Identity	Audit	Sandbox	What it is
Browser Use		—	—	🟡	🟡	Let agents drive a real browser
Skyvern		—	—	✅	🟡	Automate browser workflows with LLMs + vision
Stagehand		🟡	—	🟡	🟡	AI browser automation (Browserbase)

Voice & real-time

Framework	Stars	Guardrails	Identity	Audit	Sandbox	What it is
Pipecat		🟡	—	✅	—	Real-time voice & multimodal agents
LiveKit Agents		🟡	🟡	✅	—	Realtime voice/video agents over WebRTC
Parlant		✅	—	✅	—	Controllable, guideline-driven conversational agents

Agent infrastructure (tools · memory · sandboxes)

Project	Stars	Guardrails	Identity	Audit	Sandbox	What it is
Composio		🟡	🟡	🟡	—	Tooling/integration layer — hundreds of tools for agents
E2B		—	—	🟡	✅	Secure cloud sandboxes for AI code execution
Mem0		—	—	—	—	Memory layer for agents
Semantic Router		🟡	—	—	—	Fast routing/decision layer for agents

---

Adding an entry

1. Drop the framework in the most fitting section (keep rows roughly ordered by relevance).
2. Link the name to the GitHub repo: [Name](https://github.com/<owner>/<repo>).
3. Add the live star badge: ![](https://img.shields.io/github/stars/<owner>/<repo>?style=flat-square).
4. Rate the four governance columns ✅ / 🟡 / — per the legend; prefer ✅ only for documented, built-in support. Link every ✅ or 🟡 directly to the supporting GitHub code or documentation and leave — unlinked.

Stars are dynamic. Each badge is a shields.io query against the GitHub API (img.shields.io/github/stars/<owner>/<repo>), so counts refresh on every page view — no manual updates needed. Moved repos resolve through GitHub's redirect. Categories overlap — a framework is listed once, in its primary home.