Resource archive · retained for referenceExplore design skills →

Agent Frameworks

Frameworks for orchestrating tool use, memory, planning, and multi-step agent behavior.

LangChain

Use LangChain when one workflow needs to coordinate models, tools, and context.

Use LangGraph when an agent needs state, approvals, or retryable steps.

Use LlamaIndex when your product depends on search, documents, or private knowledge.

PydanticAI is designed for Python teams that want structured outputs and predictable agent behavior.

Use Gemini CLI when you want a terminal-first coding agent with explicit context and tool boundaries.

Use Agno when an agent has to become a managed product surface, not just a local demo.

CrewAI lets teams model agents as specialists that collaborate on a shared outcome.

AutoGen helps teams prototype agents that talk to each other and to humans.

Mastra is designed for teams that want a clean, developer-friendly stack for AI products.

Use Dify when you want one place to build and launch an AI app.