State History, Replay and Time Travel

Every step is a snapshot

Because a checkpoint is saved after each step, you can list a thread's state history (get_state_history), inspect what the state looked like before any node ran, see which node is next, and even resume from an earlier checkpoint (optionally after editing the state) to explore a different outcome. This "time travel" helps debug agents (why did it take that branch?), recover from a bad step, and test "what if" changes without re-running everything from the start. It also supports audit trails: you can show exactly what the agent knew at each step.

Listing a thread's history, run

I ran this offline with langgraph 1.2.12 and langchain-core 1.6.6 in a Python virtual environment. No API key or model is needed because plain Python functions stand in for the model, so the output is repeatable. Oldest first: the empty start (next node is the __start__ marker), then n=0 before a, n=1 before b, and the final n=2 with nothing left to run.

from typing import TypedDict
from langgraph.graph import StateGraph, START, END
from langgraph.checkpoint.memory import InMemorySaver

class State(TypedDict):
    n: int

def inc(s): return {"n": s["n"] + 1}

g = StateGraph(State)
g.add_node("a", inc); g.add_node("b", inc)
g.add_edge(START, "a"); g.add_edge("a", "b"); g.add_edge("b", END)
app = g.compile(checkpointer=InMemorySaver())
cfg = {"configurable": {"thread_id": "t"}}
app.invoke({"n": 0}, cfg)

snaps = list(app.get_state_history(cfg))          # newest first
for s in reversed(snaps):
    print("n =", s.values.get("n"), "| next:", s.next)

Output:

n = None | next: ('__start__',)
n = 0 | next: ('a',)
n = 1 | next: ('b',)
n = 2 | next: ()

Quick check: What does time travel let you do?

• Inspect and resume from earlier checkpoints of a thread
• Change the past of the real world
• Skip all nodes
• Disable checkpoints