Intermediate9 min
RemoteGraph
Running graphs as remote services. Client-server model, SDK integration, authentication, and streaming over HTTP.
Quick Reference
- →RemoteGraph lets you invoke a LangGraph deployed on LangGraph Platform as if it were a local graph — same API, remote execution
- →Initialize with RemoteGraph(graph_id, url, api_key) and call .invoke(), .stream(), or .astream() identically to a local graph
- →Supports full streaming over HTTP: token-by-token LLM output, intermediate state updates, and tool call events
- →Use RemoteGraph to compose distributed architectures where sub-agents run on separate deployments with independent scaling
- →Authentication uses API keys passed in the client constructor or LangSmith API keys for managed deployments
What Is RemoteGraph
Local API, remote execution
RemoteGraph lets you invoke a LangGraph deployed on LangGraph Platform as if it were a local graph. Same .invoke(), .stream(), .astream() API — the execution happens remotely over HTTP. Your code does not need to know whether the graph is local or remote.
This enables distributed agent architectures. A supervisor agent running locally can delegate to specialized worker agents deployed on separate infrastructure, each scaling independently. The supervisor calls remote.invoke() the same way it would call a local graph — the network boundary is invisible.