LangSmith

LangSmith is LangChain's unified platform for LLM observability, evaluation, prompt management, and deployment of agent applications. It traces requests, evaluates outputs, and monitors deployments in one place — with deep LangGraph integration, visual prototyping in Studio, and hosted deployment for production agents. Used by Klarna, Replit, Elastic, Uber, and J.P. Morgan.

Key Features

• Full Tracing: Complete agent traces including LangGraph workflows; SDKs for Python, JS/TS, and REST API.
• Evaluations: Offline and online evals, dataset management, and regression tracking.
• Prompt Management: Prompt hub with versioning and rollout.
• Monitoring: Dashboards, alerts, annotations, and cost tracking.
• LangSmith Engine: Detects issues in LangGraph agent traces and proposes fixes — open a PR directly.
• LangSmith Fleet: No-code agent builder for templates, integrations, and routine automation.
• Deployment: Deploy and scale LangGraph agents on a purpose-built platform; visual prototyping in Studio.
• Framework Support: LangChain, LangGraph, OpenAI, Anthropic, LlamaIndex, CrewAI, and more.

Use Cases

Who Should Use This Tool?

• LangGraph Teams: Teams building stateful agents who want integrated observability and deployment.
• LLM Product Teams: Anyone shipping LLM apps who needs tracing, evals, and prompt management.
• Enterprise AI: Organizations that need governed, monitored agent deployments.

Problems It Solves

1. Fragmented tooling: Tracing, evals, prompts, and deployment in one platform.
2. Agent debugging: Engine detects issues in traces and proposes concrete fixes.
3. Production deployment: Purpose-built hosting for long-running stateful agents.

Pricing

Advantages & Unique Selling Points

Compared to Competitors:

1. Deepest LangGraph integration: Observability and deployment designed around LangGraph agents.
2. Issue detection: LangSmith Engine proactively finds problems in traces and suggests fixes.
3. One platform, full lifecycle: From tracing to evals to production deployment.

What Makes It Stand Out:

• LangSmith Fleet for no-code agent building.
• Visual prototyping in LangSmith Studio.
• Trusted by Klarna, Replit, Elastic, Uber, and J.P. Morgan.

Getting Started

Quick Start Guide

1. Sign up: Create an account at smith.langchain.com.
2. Install: pip install langsmith or use the JS/TS SDK.
3. Set up tracing: Wrap your LangChain/LangGraph calls or any LLM SDK.
4. Evaluate and deploy: Run evals, manage prompts, and deploy agents.

Integration

Integrates with:

• LangChain, LangGraph, LlamaIndex, CrewAI
• OpenAI, Anthropic, and other providers
• LangSmith Deployment and Studio

Frequently Asked Questions

Do I need LangChain to use LangSmith?

No — LangSmith supports OpenAI, Anthropic, LlamaIndex, CrewAI, and other frameworks directly.

Can it deploy agents?

Yes — LangSmith Deployment hosts and scales LangGraph agents on a purpose-built platform.

Is there a free tier?

Yes — individuals can start with a generous free tier.

Alternatives

If LangSmith isn't the right fit, consider these alternatives:

• Langfuse: Open-source self-hostable observability with 40+ integrations.
• Helicone: LLM observability with gateway caching and rate limiting.
• AgentOps: Agent-focused session recording and replay.

Tips & Best Practices

1. Use Engine to find issues: Let it scan traces and open fix PRs automatically.
2. Track regressions with evals: Run offline evals on dataset changes before shipping.
3. Prototype in Studio: Visualize and iterate on agent graphs before deployment.

Conclusion

LangSmith is the unified platform for the full agent lifecycle — trace, evaluate, manage prompts, and deploy, all integrated around LangGraph. If you build agents with LangChain/LangGraph, LangSmith is the natural observability and deployment layer.