CVE-2026-48121 is a medium-severity (CVSS 6.7) CWE-943 vulnerability in @langchain/langgraph-checkpoint-mongodb. A fix is available for @langchain/langgraph-checkpoint-mongodb — see the affected versions and patch details below.
@langchain/langgraph-checkpoint-mongodb: NoSQL parameter injection in MongoDBSaver allows cross-tenant state access
Exploitation Status
No confirmed exploitation observed yet
- A successful exploit gives an attacker total control of the affected component, not partial access.
- CISA’s own triage has not observed active exploitation or public proof-of-concept code for this CVE as of its last assessment.
Exploitation and automatability from CISA’s SSVC triage for CVE-2026-48121.
EPSS Exploitation Probability
Probability of exploitation in the next 30 days, from FIRST.org EPSS.
How urgent is this, really
CVE-2026-48121 by exploitation likelihood (EPSS) against impact (CVSS). Outside the shaded patch-first corner.
Where this sits among everything scored
Of 394,803 CVEs with a current EPSS score, this one falls in the < 10% band (highlighted). Counts from FIRST.org, log-scaled.
Real-World Exposure
How broadly this vulnerability is actually deployed: weekly install volume shows current usage, and reverse-dependency count shows how many other packages break if it stays unpatched.
@langchain/langgraph-checkpoint-mongodbnpmDescription
Summary
A NoSQL injection vulnerability existed in MongoDBSaver where checkpoint identifier fields from config.configurable were used in MongoDB queries without strict type enforcement. In vulnerable versions, attacker-controlled object payloads (for example MongoDB operators like $gt and $ne) could be interpreted as query operators instead of literal identifier values.
This could bypass intended thread scoping and return checkpoints from other tenants.
Attack surface
The vulnerable path was in MongoDBSaver.getTuple(), where thread_id, checkpoint_ns, and checkpoint_id were used in MongoDB find() queries. The same unvalidated values were then reused to fetch pending writes.
Applications were exposed when untrusted input was forwarded into config.configurable (for example, directly from request bodies or query parameters) without string coercion or schema validation.
Who is affected?
Applications are vulnerable if they:
- Use
@langchain/langgraph-checkpoint-mongodbwith multi-tenant or user-isolated thread models. - Accept user-controlled values for
thread_id,checkpoint_ns, orcheckpoint_id. - Pass those values into
app.invoke(),app.stream(), or direct saver methods without validation.
Applications are generally not vulnerable if they:
- Use server-issued identifiers only.
- Source
thread_idfrom trusted URL params that remain strings. - Enforce schema validation that rejects non-string identifier fields.
Impact
An attacker with control over configurable checkpoint identifiers could read checkpoint data outside their authorized thread boundary.
Potentially exposed data includes:
- Checkpoint state
- Metadata
- Pending writes
This is a confidentiality issue with cross-tenant data disclosure risk.
Exploit example
An attacker-controlled request can inject MongoDB operators:
graph = new StateGraph(...)
.compile({
checkpointer: new MongoDBSaver()
});
graph.invoke(..., {
configurable: {
"thread_id": { "$gt": "" },
"checkpoint_ns": { "$ne": null }
}
});
If this payload is forwarded into config.configurable, the resulting query may match checkpoints outside the intended tenant/thread scope.
Security hardening changes
Version 1.3.1 hardens @langchain/langgraph-checkpoint-mongodb by adding runtime validation for configurable checkpoint identifiers and rejecting invalid values before MongoDB query/write paths execute.
The patch also includes regression tests covering object/operator payloads across affected methods.
Migration guide
Upgrade to @langchain/[email protected] or later.
No API migration is required for valid callers. However, applications that currently pass non-string identifier values in config.configurable will now receive explicit errors and should normalize/validate inputs.
As defense in depth, validate identifier fields at API boundaries and avoid passing raw client objects into graph config.
Resources
Affected Packages
| Ecosystem | Package | Vulnerable range | Fix |
|---|---|---|---|
| 📦npm | @langchain/langgraph-checkpoint-mongodb | all versions | 1.3.1npm install @langchain/langgraph-checkpoint-mongodb@1.3.1 |
Detection & mitigation playbook
Open-source dependencyDetect
Scan your dependency tree (package-lock.json, pnpm-lock.yaml, requirements.txt, go.sum, etc.) for @langchain/langgraph-checkpoint-mongodb, including transitive dependencies — a direct dependency you never call can still pull in a vulnerable version.
Fix
Update @langchain/langgraph-checkpoint-mongodb to 1.3.1 or later, then make sure no transitive (indirect) dependency still pins the vulnerable range — O3 confirms CVE-2026-48121 is resolved across your whole dependency graph.
Workarounds
Until you can upgrade, make sure every query built from user input uses parameterised statements or a prepared-statement API rather than string concatenation, and reduce the database account's privileges so an injected query cannot read or alter data beyond what the feature needs.
Frequently Asked Questions
Is CVE-2026-48121 in your dependencies?
Find it across npm, including transitive dependencies.