★ OverviewIntermediate12 min
LangSmith Deployment (LangGraph Platform)
LangGraph Platform is now LangSmith Deployment — a managed hosting platform for long-running, stateful agents. Cloud, self-hosted, standalone server, and hybrid options.
Quick Reference
- →LangGraph Platform is now called LangSmith Deployment — deploy and scale agents from the same platform as observability
- →Deploy with a single langgraph.json config file that declares graphs, dependencies, environment variables, and Python/Node version
- →Agent Server handles checkpointing, persistence, and store automatically — no manual configuration needed
- →LangGraph Studio (in LangSmith UI) provides visual debugging, breakpoints, state inspection, and thread replay
- →Built-in support for streaming, background runs, cron jobs, webhook triggers, and thread-level persistence out of the box
Deployment Modes
LangGraph Platform → LangSmith Deployment
LangGraph Platform has been rebranded as LangSmith Deployment. It is now integrated into LangSmith — the same platform you use for tracing and evaluation. All deployment management (Cloud, self-hosted, standalone) is accessed through LangSmith.
| Mode | Hosting | Persistence | Best For |
|---|---|---|---|
| Cloud (LangSmith managed) | LangChain-managed infrastructure | Built-in — Agent Server handles it | Prototyping, startups, fast iteration |
| Self-Hosted Enterprise | Your Kubernetes / Docker infrastructure | You configure | Full control, data residency, compliance |
| Standalone Server | Your own server (without full LangSmith UI) | You configure | Lightweight self-hosting, no UI dependency |
| Hybrid | Data plane in your cloud, control plane on LangSmith | Managed in your cloud | Compliance + cloud convenience |