Key Features

• LangGraph (open-source agent orchestration): A framework for building stateful, cyclical agent workflows with first-class support for multi-agent architectures, human-in-the-loop checkpoints, persistent memory, and parallel sub-agent execution — the structural backbone for most production-grade agents built on LangChain.
• LangSmith Observability (tracing and debugging): Framework-agnostic tracing that records every step of every agent run — LLM calls, tool invocations, retrieval steps, sub-agent delegation — with an AI assistant (Polly) for navigating and debugging large traces.
• LangSmith Evaluation: Offline dataset-based and online production evaluations with LLM-as-judge scoring, annotation queues for human review, and pairwise comparison — allowing teams to measure agent quality systematically rather than relying on manual spot-checking.
• LangSmith Deployment (managed agent hosting): A production runtime for LangGraph agents offering durable execution, horizontal autoscaling, human-in-the-loop interrupts, versioning, rollbacks, and native support for A2A, MCP, and Agent Protocol standards.
• LangSmith Fleet (no-code agent deployment): Formerly Agent Builder; a natural-language interface allowing non-technical teams to describe an agent task (daily briefings, competitor tracking, project updates) and have LangSmith build and deploy the agent — with human approval before sensitive actions.
• 1,000+ model and tool integrations: Native integrations with every major LLM provider, vector database, search tool, and business API — keeping teams model-agnostic and allowing migration between providers without rewriting agent logic.