LangSmith
LangChain's platform for tracing, monitoring, evaluating and debugging LLM apps and agents in production
LangSmith is LangChain's observability, evaluation and deployment platform for AI agents. It records what an agent does step by step, including LLM calls, tool calls and multi-turn conversations, so developers can find failures and track latency, cost and quality, whatever framework the agent uses.
Key features:
- Trace agents natively from popular frameworks or OpenTelemetry.
- Monitor cost, tool use and agent trajectories, with webhook and PagerDuty alerts.
- Score quality with online LLM-as-judge and code evals, datasets and annotation queues.
- Cluster traces automatically with Insights to surface usage patterns and failure modes.
- Iterate on prompts in Prompt Hub and the Playground.
- Deploy long-running agents with LangSmith Deployment and expose them as MCP servers.
SDKs are available for Python, TypeScript, Go and Java. The Developer plan is free for one seat with 5,000 traces a month; Plus is $39 per seat a month with 10,000 traces, then pay as you go, and Enterprise adds self-hosted and hybrid deployment, SSO and RBAC.