langgraph
by langchain-ai
Compose resilient language agents as explicit graphs for safer delegation
What It Does
Models agent behavior as composable directed graphs to build resilient language agents. Uses graph nodes for prompts, tools, and control flow so developers can reason about delegation, retries, and fallbacks. Supports Python SDK with integrations to common LLM providers and LangChain components for prompt/tool wiring. Hierarchical Multi-Agent Pattern and A2A Protocol Pattern help structure these capabilities across complex agent setups.
Key Benefits
When to Use
Developers building multi-step or multi-agent language workflows who need explicit control flow, retries, and observability in production. For practical workflow orchestration and planning, consider the Planning Pattern to design robust retry and fall-back strategies.
How It's Used
- Modeling agent delegation chains with explicit nodes and fallbacks
- Adding retries, timeouts, and conditional logic to complex prompt pipelines
- Testing and debugging multi-step agent workflows before production
- Integrating LLMs and tools in reproducible, auditable control graphs