🧩 Build agents (frameworks)
SDKs, frameworks and visual builders for building your own AI agents.
How we rank: editorial order informed by public signals shown on each card (GitHub stars for open-source tools, notable adoption otherwise). Rankings are never for sale; paid placements appear only in the labeled Featured slot.
Dify
FreeOpen-source platform for building agentic workflows and LLM apps.
Why people use it: Visual builder + RAG + agents, self-hostable with one Docker command.
LangChain
FreeThe widely used framework for building LLM apps and agents.
Why people use it: Huge ecosystem of integrations and examples; most tutorials start here.
Ollama
FreemiumRun open-weight LLMs locally with one command.
Why people use it: The easy way to power agents with local models: no API bill, works offline.
AutoGen
FreeMicrosoft's framework for multi-agent AI applications.
Why people use it: Research-grade multi-agent conversations from Microsoft.
CrewAI
FreeFramework for orchestrating role-playing, collaborating AI agents.
Why people use it: Define a 'crew' of agents with roles and goals in a few lines of Python.
Flowise
FreeDrag-and-drop UI to build AI agents and LLM flows.
Why people use it: Low-code canvas for agents, popular with makers who don't want to write Python.
LlamaIndex
FreeFramework for agents and RAG over your own data.
Why people use it: Go-to for agents that need to read your documents and data.
LangGraph
FreeLow-level orchestration for stateful, long-running agents.
Why people use it: When you need control: explicit graphs, persistence and human-in-the-loop.
OpenAI Agents SDK
FreeLightweight Python framework for multi-agent workflows.
Why people use it: Minimal primitives (agents, handoffs, guardrails) from OpenAI.
Mastra
FreeTypeScript framework for AI agents, workflows and RAG.
Why people use it: The TypeScript-native option for web devs building agents.
Vercel AI SDK
FreeTypeScript toolkit for building AI apps and agents.
Why people use it: Streaming chat UIs and tool-calling agents in Next.js with any model provider.
Pydantic AI
FreeType-safe Python agent framework from the Pydantic team.
Why people use it: Structured, validated outputs and FastAPI-style ergonomics for agents.
Pi
FreeMIT-licensed, extensible agent harness; 1.0 shipped Oct 1, 2026.
Why people use it: New: a minimal harness you extend yourself instead of a big framework.
Agent-Native (beta)
FreeOpen-source TypeScript framework for agents that ship with their own UI.
Why people use it: New: define an action once and expose it to UI, API, CLI, MCP and A2A.
n8n
FreemiumFair-code workflow automation platform with native AI agent nodes.
Why people use it: No-code way to wire up agents with tools and memory.