LangChain

LangChain provides an agent engineering platform, LangSmith, alongside the open-source frameworks LangChain, LangGraph and deepagents. It is aimed at teams building AI agents, from developers writing code to non-technical users working with no-code agents. LangSmith covers the agent lifecycle: building agents with control over context, memory and model choice, tracing and debugging runs, running evaluations, deploying agents with isolated sandboxes and an LLM gateway, monitoring production with online evals, dashboards and alerts, and governance features such as cost limits, sensitive-data detection and human approval for sensitive actions.

• Open-source agent frameworks: LangChain, LangGraph and deepagents
• Step-by-step tracing and debugging of agent runs, with evaluations built from expert review
• Agent deployment with isolated sandboxes for agent-generated code and an LLM gateway
• Production monitoring with online evals, dashboards and alerts
• Governance: cost limits, sensitive-data detection and human approval for sensitive agent actions