What it does
LangChain presents itself as an open agent platform that covers the whole agent development lifecycle: build, test, deploy, monitor, iterate and govern. On the open source side it offers three frameworks. LangChain is for starting agents quickly with any model provider, LangGraph gives low-level control for production agents that need some determinism, and deepagents targets long-running agents handling open-ended work. Alongside the frameworks sits the LangSmith platform, which adds observability, evaluation, deployment, sandboxes for running agent-generated code, an LLM Gateway for controlling model calls, and Fleet for building agents without code. The site says deployment can be SaaS, BYOC or self-hosted.
Capabilities
- Open source frameworks: LangChain, LangGraph and deepagents
- Step-by-step tracing and observability for agent runs
- Evaluation built from traces and expert review
- Deployment for long-running agents with persistent state
- Sandboxes for running agent-generated code in isolation
- LLM Gateway to control model calls, spend and sensitive data
When to use it
- Building agents on top of any model provider with an open source framework
- Putting production agents under tracing, evals and monitoring
- Deploying long-running agents in SaaS, BYOC or self-hosted environments
- Letting non-engineering teams create agents without writing code
