# Agent Frameworks

> AI agent frameworks for building systems where a model plans, calls tools, and works across multiple steps: multi-agent frameworks like AutoGen and CrewAI, code-first orchestration SDKs, and higher-level application frameworks like Dify. Building your first agent? Start with an application framework; reach for an orchestration SDK when you need control over every step.

36 tools in 3 subcategories. Web page: https://ailandscape.org/category/agent-frameworks · Each tool also has a markdown version at /tool/{slug}.md

## Orchestration SDKs

- [agent-orchestrator](https://ailandscape.org/tool/agent-orchestrator): Per-agent worktree and branch isolation; auto-fixes CI failures and opens PRs
- [AgentForge](https://ailandscape.org/tool/agentforge): Low-code framework for building, testing, and iterating on AI agents
- [agentscope](https://ailandscape.org/tool/agentscope): Alibaba's agent-oriented programming framework for building LLM applications
- [antfarm](https://ailandscape.org/tool/antfarm): Build your agent team with one command
- [ChatDev](https://ailandscape.org/tool/chatdev): Communicative LLM agents simulating a virtual software development company
- [claude-flow](https://ailandscape.org/tool/claude-flow): Deploy multi-agent swarms with coordinated, programmable workflows
- [clawe](https://ailandscape.org/tool/clawe): Trello-style multi-agent coordination system
- [Daydreams](https://ailandscape.org/tool/daydreams): Open-source framework for building and orchestrating autonomous AI agents
- [FastAgency](https://ailandscape.org/tool/fastagency): Rapid development framework for multi-agent AI applications
- [gastown](https://ailandscape.org/tool/gastown): Multi-agent orchestration system with persistent work tracking across sessions
- [Google ADK](https://ailandscape.org/tool/google-adk): Google's open-source Agent Development Kit
- [humanlayer](https://ailandscape.org/tool/humanlayer): SDK for injecting human approval gates into agent loops
- [L2MAC](https://ailandscape.org/tool/l2mac): LLM Automatic Computer for generating extensive codebases and complex artifacts
- [lalph](https://ailandscape.org/tool/lalph): LLM agent orchestrator that connects to your issue tracker and spawns agents
- [LangGraph](https://ailandscape.org/tool/langgraph): Graph-based agent workflow orchestration
- [Langroid](https://ailandscape.org/tool/langroid): Lightweight, extensible Python framework for multi-agent LLM programming
- [loom](https://ailandscape.org/tool/loom): Autonomous development loops that iteratively evolve a software product
- [OpenAI Agents SDK](https://ailandscape.org/tool/openai-agents-sdk): OpenAI's official Python SDK for orchestrated multi-agent systems with handoffs
- [openfang](https://ailandscape.org/tool/openfang): Open-source Agent Operating System
- [opengoat](https://ailandscape.org/tool/opengoat): Build autonomous organizations of AI agents
- [paperclip](https://ailandscape.org/tool/paperclip): Orchestration infrastructure for zero-human-intervention companies
- [PydanticAI](https://ailandscape.org/tool/pydanticai): Production-grade, type-safe Python agent framework by Pydantic
- [Questflow](https://ailandscape.org/tool/questflow): Decentralized multi-agent orchestration with a marketplace for deploying agents
- [smolagents](https://ailandscape.org/tool/smolagents): Hugging Face's minimal library for code-first agents
- [Strands Agents](https://ailandscape.org/tool/strands-agents): AWS-backed model-driven agent SDK for production

## Multi-Agent Frameworks

- [AutoGen](https://ailandscape.org/tool/autogen): Multi-agent conversation framework by Microsoft
- [AutoGPT](https://ailandscape.org/tool/autogpt): Autonomous AI agent platform
- [BabyAGI](https://ailandscape.org/tool/babyagi): Minimal task-driven autonomous agent using GPT-4 and vector databases
- [CAMEL](https://ailandscape.org/tool/camel): First LLM multi-agent framework and community dedicated to agent scaling laws
- [CrewAI](https://ailandscape.org/tool/crewai): Framework for orchestrating role-playing AI agents
- [GPT Engineer](https://ailandscape.org/tool/gpt-engineer): Describe what to build; AI asks clarifying questions and generates the codebase
- [LoopGPT](https://ailandscape.org/tool/loopgpt): AutoGPT re-implementation with modular architecture and better memory handling
- [Mastra](https://ailandscape.org/tool/mastra): TypeScript AI framework with workflows, agents, RAG, integrations, and evals
- [MetaGPT](https://ailandscape.org/tool/metagpt): Assigns roles to LLMs to form a collaborative multi-agent software entity

## LLM Application Frameworks

- [Agno](https://ailandscape.org/tool/agno): Full-stack framework for building multi-agent systems with memory
- [Dify](https://ailandscape.org/tool/dify): Open-source LLM app platform with visual workflow editor, RAG, and agent builder

---

Part of [AI Landscape](https://ailandscape.org), an open map of the AI ecosystem. Index for AI assistants: https://ailandscape.org/llms.txt