displayName
LangGraph Graph Core (current)
agentVersionId
packageRef
loopIteratorPolicy
custom
loopIteratorNotes
The core loop is an explicit graph/state-machine authored by the application:
nodes transform state, edges route control, and terminal conditions are part
of the graph definition rather than a fixed CLI loop.
contextManagementStrategy
user-managed
compactionTriggerNotes
Context and memory are represented in graph state, stores, checkpointers, or
model-call middleware chosen by the graph author.
subagentInvokerPolicy
nested-loop
subagentInvokerNotes
Custom multi-agent patterns can be encoded as subgraphs, supervisor graphs,
handoff nodes, or Deep Agents-style subagent harnesses.
resultSynthesisPolicy
post-processed
resultSynthesisNotes
Final outputs are graph-defined: the terminal state can be converted to chat
messages, structured outputs, events, or application-specific payloads.
stopDetectionStrategy
structured-end-event
transportClientLibrary
langchain-chat-model-abstraction
supportedTransportProtocols
parallelToolCallHandling
native
streamingFidelity
full
thinkingChannelHandling
model-dependent
notes
Represents LangGraph's framework-level core, not a single product loop. The
graph author chooses which model provider integrations and node semantics are
active for a concrete application.