LangGraph API

OpenMemory implements the LangGraph CheckpointSaver interface, allowing you to persist the state of your stateful graphs (agents) automatically.

Python Usage

from langgraph.checkpoint.openmemory import OpenMemorySaver
from openmemory import OpenMemory
from langgraph.graph import StateGraph

# 1. Initialize Memory
mem = OpenMemory()

# 2. Create Checkpointer
checkpointer = OpenMemorySaver(mem)

# 3. Compile Graph with Checkpointer
workflow = StateGraph(...)
# ... define nodes ...
app = workflow.compile(checkpointer=checkpointer)

# 4. Run with Thread ID
config = {"configurable": {"thread_id": "thread_1"}}
app.invoke({"messages": ["Hi"]}, config=config)

Features

State Persistence

Every step of your agent's execution is saved to OpenMemory. If the process crashes, you can resume exactly where you left off.

Time Travel

You can query past states of the agent using OpenMemory's temporal graph features.

# Get state from 5 minutes ago
history = mem.get_state_history("thread_1", time_delta="-5m")

Human-in-the-loop

Because state is persisted, you can pause execution, wait for human approval, and then resume the graph using the same thread_id.