Build an AI agent in code

▸ Largest ecosystem; LangGraph for controllable agents.

Open-source frameworks for LLM apps and stateful agents, plus the LangSmith platform.

▸ Minimal primitives: agents, handoffs, guardrails, tracing.

Lightweight open-source SDK for multi-agent workflows.

▸ Reuse the Claude Code agent loop, tools and context management.

SDK exposing the Claude Code agent harness for custom agents.

▸ Typed, production-friendly Python agents.

Type-safe Python agent framework from the Pydantic team.

▸ Top choice for TypeScript developers.

TypeScript agent framework with memory, tools, MCP and observability.

▸ Unified provider API and streaming UI for web apps.

TypeScript toolkit for AI apps and agents.

▸ Model-agnostic, deploys easily to Google Cloud.

Google's open-source Agent Development Kit.

▸ Simple role-based multi-agent orchestration.

Framework and platform for multi-agent 'crews'.

▸ Agent workflows over your data.

Agent framework plus LlamaParse document parsing and workflows.

▸ Fast, lightweight agents with built-in memory.

Python framework and runtime for multi-agent systems.

▸ Tiny codebase; agents that write Python actions.

Hugging Face's minimal code-acting agent library.

▸ Influential research multi-agent framework; check project status before new work.

Microsoft's open-source multi-agent framework.

▸ Free way to prototype Gemini prompts and grab API code.

Google's studio for Gemini models, including 'Build' vibe-coding mode.

▸ Not a framework: swap models behind one API.

Unified API for hundreds of LLMs.