CVE-2026-115883: LangGraph: Namespace labels containing `.` reach other namespaces in Postgres and SQLite stores
The LangGraph Postgres and SQLite stores let a namespace label containing a dot reach another namespace. A caller who controls a label can read, overwrite or delete another tenant's items. Upgrade langgraph-checkpoint-postgres to 3.1.3 and langgraph-checkpoint-sqlite to 3.1.2. InMemoryStore is not affected.
Published Oct 23, 2026 · Updated Oct 25, 2026 · Analysis Nov 2, 2026
What an attacker can do
These stores save a namespace as its labels joined with a dot. So ("tenant", "bob") and ("tenant.bob",) map to the same stored namespace. put rejected labels with a dot, but get, search, delete, list_namespaces, their async versions and ops passed to batch/abatch did not. That includes PutOp items inside a batch.
In practice, get(("tenant.bob",), key) returned data stored for ("tenant", "bob"). delete(("tenant.alice",), key) removed Alice's item. A batch of PutOp overwrote Bob's item.
You are exposed if you use one of these stores, rely on namespaces to separate users or tenants, and let untrusted input pick a label that reaches a read, delete or batch call. On LangSmith Deployment, the HTTP store API already rejected such labels, but store calls from graph code reached the flaw.
Who is affected
langgraph-checkpoint-postgres before 3.1.3 and langgraph-checkpoint-sqlite before 3.1.2, including the async store variants. LangSmith Agent Server is fixed from 0.15.5 and 0.16.2. Check installed versions:
pip show langgraph-checkpoint-postgres langgraph-checkpoint-sqlitelanggraph-checkpoint-postgres
- Ecosystem
- PyPI
- Affected
- < 3.1.3
- Fixed in
- 3.1.3
- Check
pip show langgraph-checkpoint-postgres
langgraph-checkpoint-sqlite
- Ecosystem
- PyPI
- Affected
- < 3.1.2
- Fixed in
- 3.1.2
- Check
pip show langgraph-checkpoint-sqlite
How to fix it
Upgrade to langgraph-checkpoint-postgres 3.1.3 or langgraph-checkpoint-sqlite 3.1.2. Both require langgraph-checkpoint 4.3.0:
pip install -U "langgraph-checkpoint-postgres>=3.1.3" "langgraph-checkpoint-sqlite>=3.1.2"The fixed code raises InvalidNamespaceError for labels that are empty, not strings, or contain a dot. Data written through put never contained dots, so nothing becomes unreachable. LangSmith Cloud deployments get the fix on their next revision. Self-hosted deployments must upgrade. Custom stores that join namespaces into text can call langgraph.store.base.validate_op_namespace on each op.
How to tell if you were hit
Search application code for places where user input becomes a namespace label in get, search, delete, list_namespaces or batch. In logs, look for namespace labels containing a . in those calls, and for reads or deletes of a tenant's keys by a different user. After upgrading, such attempts raise InvalidNamespaceError, which is worth alerting on.
Original advisory
From the GitHub Advisory Database (CC BY 4.0).
Summary
The Postgres and SQLite stores save a namespace as its labels joined with ., so ("tenant", "bob") and ("tenant.bob",) map to the same stored namespace. put rejected labels containing ., but get, search, delete, list_namespaces, their async versions, and ops passed directly to batch/abatch did not. A caller that could pass such a label could read, overwrite or delete items in another namespace.
Details
BaseStore checked namespace labels only in put/aput. The other methods passed the namespace on unchecked, and batch/abatch checked nothing, including the PutOp items inside them. PostgresStore, AsyncPostgresStore, SqliteStore and AsyncSqliteStore turn a namespace into text with ".".join(namespace), so a label containing . resolved to the namespace whose labels it spells out:
get(("tenant.bob",), key),search(("tenant.bob",))andlist_namespaces(prefix=("tenant.bob",))returned data stored under("tenant", "bob").delete(("tenant.alice",), key)deleted the item stored under("tenant", "alice").batch([PutOp(("tenant.bob",), key, value)])overwrote the item stored under("tenant", "bob").
All releases of langgraph-checkpoint-postgres before 3.1.3 and langgraph-checkpoint-sqlite before 3.1.2 are affected. InMemoryStore compares namespaces as tuples and is not affected.
Impact
You may be affected if you use one of these stores, rely on namespaces to keep users or tenants apart, and let untrusted input choose a namespace label that reaches a read, delete or batch call. Such input could read, overwrite or delete another tenant's items. Labels written through put could never contain ., so no stored data becomes unreachable after upgrading.
LangSmith Deployment: the HTTP store API already rejected such labels, but store calls made from graph code reached the same flaw. The Agent Server rejects these labels from versions 0.15.5 and 0.16.2. On LangSmith Cloud, new deployments and revisions already include the fix and existing deployments get it on their next revision; self-hosted deployments should upgrade.
Remediation
Upgrade to langgraph-checkpoint-postgres 3.1.3 or langgraph-checkpoint-sqlite 3.1.2. Both require langgraph-checkpoint 4.3.0, where get, search, delete and list_namespaces reject labels that are empty, are not strings, or contain . with InvalidNamespaceError, and the Postgres and SQLite stores apply the same check to every op passed to batch/abatch. Stores outside this repository that turn namespaces into delimited text can call langgraph.store.base.validate_op_namespace for each op they run.
Credited to VuxNx, oct0pus7.
Questions
- Does InMemoryStore have this problem?
No. InMemoryStore compares namespaces as tuples, so a dotted label cannot collide with a multi-label namespace.
- Can an attacker exploit this with no control over namespace labels?
No. The caller must be able to pass a dotted label to a read, delete or batch call. If your code builds namespaces only from trusted values, you are not exposed.
- Will existing data break after upgrading?
No. Labels written through put could never contain a dot, so no stored data becomes unreachable.
References
Read for this analysis:
- https://github.com/langchain-ai/langgraph/security/advisories/GHSA-6hmw-9432-67qq
- https://github.com/langchain-ai/langgraph/pull/9217
- https://github.com/langchain-ai/langgraph/commit/bfcfea554ed5c7f7be562cebf8825e911b493ab1
- https://github.com/langchain-ai/langgraph/releases/tag/checkpointpostgres%3D%3D3.1.3
- https://github.com/langchain-ai/langgraph/releases/tag/checkpointsqlite%3D%3D3.1.2
How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. The analysis is written by an AI model from the sources listed above and checked by automated rules before it is published. Check versions against the vendor's advisory before you act. Methodology
More PyPI vulnerabilities
All PyPI entriesCVE-2026-71433
LangGraph: Namespace prefix matching crosses segment boundaries in Postgres and SQLite stores
medium5.3Aug 6, 2026CVE-2025-67644
LangGraph's SQLite is vulnerable to SQL injection via metadata filter key in SQLite checkpointer list method
high7.3Dec 10, 2025CVE-2025-64104
LangGraph SQLite Checkpoint Filter Key SQL Injection POC for SqliteStore
high7.3Oct 29, 2025CVE-2025-8709
LangGraph's SQLite store implementation has a SQL Injection Vulnerability
high7.3Oct 26, 2025
