Implementation:Langchain ai Langgraph StateGraph Add Node

Metadata	Value
Type	Implementation (API Doc)
Library	langgraph
Source File	libs/langgraph/langgraph/graph/state.py
Lines	L569-783 (add_node), L889-934 (add_sequence)
Workflow	Building_a_Stateful_Graph

Overview

StateGraph.add_node registers a processing function or runnable as a named node in the graph builder. StateGraph.add_sequence is a convenience method that registers multiple nodes and connects them with sequential edges in one call.

Description

add_node accepts either a callable/runnable directly (with the name inferred from __name__ or get_name()) or an explicit string name paired with an action. The method:

1. Validates the name is unique, not reserved (START/END), and does not contain forbidden separator characters.
2. Coerces the action into a Runnable via coerce_to_runnable().
3. Inspects the action's type hints to infer input schema and detect Command return types (for graph visualization).
4. Stores a StateNodeSpec containing the runnable, metadata, input schema, retry policy, cache policy, destinations, and defer flag.
5. If a custom input_schema is provided or inferred, registers it with _add_schema().

add_sequence iterates a list of nodes, calling add_node for each and add_edge between consecutive pairs. It validates that all names are unique.

Both methods return Self to support method chaining.

Usage

from typing_extensions import TypedDict
from langgraph.graph import StateGraph, START

class State(TypedDict):
    x: int

def step_one(state: State) -> dict:
    return {"x": state["x"] + 1}

def step_two(state: State) -> dict:
    return {"x": state["x"] * 2}

builder = StateGraph(State)
builder.add_node(step_one)                # inferred name: "step_one"
builder.add_node("custom_name", step_two) # explicit name
builder.add_edge(START, "step_one")
builder.add_edge("step_one", "custom_name")

Code Reference

Source Location

Item	Path	Lines
add_node (implementation)	libs/langgraph/langgraph/graph/state.py	L569-783
add_node (overloads)	libs/langgraph/langgraph/graph/state.py	L289-567
add_sequence	libs/langgraph/langgraph/graph/state.py	L889-934

Signature

def add_node(
    self,
    node: str | StateNode[NodeInputT, ContextT],
    action: StateNode[NodeInputT, ContextT] | None = None,
    *,
    defer: bool = False,
    metadata: dict[str, Any] | None = None,
    input_schema: type[NodeInputT] | None = None,
    retry_policy: RetryPolicy | Sequence[RetryPolicy] | None = None,
    cache_policy: CachePolicy | None = None,
    destinations: dict[str, str] | tuple[str, ...] | None = None,
) -> Self:

def add_sequence(
    self,
    nodes: Sequence[
        StateNode[NodeInputT, ContextT]
        | tuple[str, StateNode[NodeInputT, ContextT]]
    ],
) -> Self:

Import

from langgraph.graph import StateGraph

I/O Contract

add_node Parameters

Parameter	Type	Default	Description
node	StateNode	required	The function/runnable to register, or a string name (when action is also provided).
action	None	None	The function/runnable when node is a string name.
defer	bool	False	If True, defer execution until the run is about to end.
metadata	None	None	Arbitrary metadata to attach to the node.
input_schema	None	None	Custom input schema for this node. Defaults to the graph's state schema.
retry_policy	Sequence[RetryPolicy] | None	None	Retry configuration. If a sequence, the first matching policy is applied.
cache_policy	None	None	Caching configuration for this node.
destinations	tuple[str, ...] | None	None	Rendering hint for Command-returning nodes. Does not affect execution.

Returns: Self -- The StateGraph instance, for method chaining.

Raises:

• ValueError -- If the node name already exists, is reserved (START/END), or contains forbidden characters.

add_sequence Parameters

Parameter	Type	Description
nodes	tuple[str, StateNode]]	Ordered sequence of nodes or (name, node) tuples.

Returns: Self -- The StateGraph instance.

Raises:

• ValueError -- If the sequence is empty or contains duplicate names.

Usage Examples

Inferred Node Name

from typing_extensions import TypedDict
from langgraph.graph import StateGraph, START

class State(TypedDict):
    x: int

def my_node(state: State) -> dict:
    return {"x": state["x"] + 1}

builder = StateGraph(State)
builder.add_node(my_node)  # name inferred as "my_node"
builder.add_edge(START, "my_node")
graph = builder.compile()
graph.invoke({"x": 1})
# {'x': 2}

Explicit Node Name

builder = StateGraph(State)
builder.add_node("my_fair_node", my_node)
builder.add_edge(START, "my_fair_node")
graph = builder.compile()
graph.invoke({"x": 1})
# {'x': 2}

Custom Input Schema

class NodeInput(TypedDict):
    x: int

def my_node(state: NodeInput) -> dict:
    return {"x": state["x"] + 1}

builder = StateGraph(State)
builder.add_node("my_node", my_node, input_schema=NodeInput)
builder.add_edge(START, "my_node")
graph = builder.compile()
graph.invoke({"x": 1})
# {'x': 2}

Using add_sequence

def step_a(state: State) -> dict:
    return {"x": state["x"] + 1}

def step_b(state: State) -> dict:
    return {"x": state["x"] * 2}

def step_c(state: State) -> dict:
    return {"x": state["x"] - 3}

builder = StateGraph(State)
builder.add_sequence([step_a, step_b, step_c])
builder.add_edge(START, "step_a")
builder.set_finish_point("step_c")
graph = builder.compile()
graph.invoke({"x": 5})
# step_a: 5+1=6, step_b: 6*2=12, step_c: 12-3=9 -> {'x': 9}

Deferred Node

def cleanup(state: State) -> dict:
    return {"x": 0}

builder = StateGraph(State)
builder.add_node("main", my_node)
builder.add_node("cleanup", cleanup, defer=True)
builder.add_edge(START, "main")
builder.add_edge("main", "cleanup")
graph = builder.compile()

Related Pages

• Langchain_ai_Langgraph_Node_Registration
• Langchain_ai_Langgraph_StateGraph_Init
• Langchain_ai_Langgraph_StateGraph_Add_Edge
• Langchain_ai_Langgraph_StateGraph_Compile