StoreBackend: files shared across threads

StoreBackend is a backend that keeps the agent's files in a LangGraph store under a namespace, so a file written on one thread can be read on any other thread that uses the same store.

Last updated: 29 Sep, 2026 · Deep Agents 0.7

The video's two-thread test

The third backend creates a persistent key-value store. The video builds an InMemoryStore, passes it as store=, and gives StoreBackend a namespace, a function that returns ("demo-user",). On thread one it writes /notes/todo.txt; on a second thread, with a new id, it asks for the file back and gets it. With an InMemoryStore the file is not on disk at all: it lives in memory, inside the store object, as an entry under the namespace. The video's code follows, run on Groq with the model line swapped, plus a loop that lists the store's keys:

import uuid

from deepagents import create_deep_agent
from deepagents.backends import StoreBackend
from langchain.chat_models import init_chat_model
from langgraph.store.memory import InMemoryStore

store = InMemoryStore()
agent = create_deep_agent(
    model=init_chat_model("groq:openai/gpt-oss-120b", max_retries=6),
    backend=StoreBackend(namespace=lambda rt: ("demo-user",)),
    store=store,
)

thread_1 = {"configurable": {"thread_id": str(uuid.uuid4())}}
result = agent.invoke({"messages": [{"role": "user", "content": (
    "Create a file at /notes/todo.txt with exactly this content:\n"
    "1. Record video\n2. Edit video\n3. Upload video\n"
    "Then tell me you've done it.")}]}, config=thread_1)
print("--- Agent reply (thread 1) ---")
print(result["messages"][-1].content)

thread_2 = {"configurable": {"thread_id": str(uuid.uuid4())}}
followup = agent.invoke({"messages": [{"role": "user", "content": "Read /notes/todo.txt back to me verbatim."}]}, config=thread_2)
print("--- Read-back on a different thread ---")
print(followup["messages"][-1].content)

print("--- In the store ---")
for item in store.search(("demo-user",)):
    print(item.key)

--- Agent reply (thread 1) ---
The file **/notes/todo.txt** has been created with the requested content.
--- Read-back on a different thread ---
Here is the exact content of **/notes/todo.txt**:

```
1. Record video
2. Edit video
3. Upload video
```
--- In the store ---
/notes/todo.txt

What the two threads and the store show

• Thread one wrote the file.
• Thread two started with an empty conversation, yet read_file found the file: the store is outside any one thread.
• The store's key is the file's path, under the demo-user namespace.

The StoreBackend call

from deepagents.backends import StoreBackend
from langgraph.store.memory import InMemoryStore

agent = create_deep_agent(
    model=model,
    backend=StoreBackend(namespace=lambda rt: ("demo-user",)),   # where in the store
    store=InMemoryStore(),                                       # the store itself
)

The namespace function receives the run's runtime, so in a real app it can return the user's id and keep each user's files apart.

The three backends compared

Where StoreBackend fits

• Per-user files that must follow a user into every new conversation.
• Agents deployed on LangSmith, where a store is provided for you.
• Shared notes several agents read and write.

Related

• Previous: FilesystemBackend: files on your disk
• Next: CompositeBackend: long-term memory in /memories/
• Reference: Backends: StoreBackend