CVE-2026-55235 — langgraph-api
MEDIUMCVE-2026-55235 is a medium-severity (CVSS 5.9) Improper Authentication vulnerability in langgraph-api. A fix is available for langgraph-api — see the affected versions and patch details below.
langgraph-api: Relative webhook targets in LangGraph Server can reach in-process routes without authentication
Exploitation Status
No confirmed exploitation observed yet
- 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-2026-55235.
EPSS Exploitation Probability
Probability of exploitation in the next 30 days, from FIRST.org EPSS.
How urgent is this, really
CVE-2026-55235 by exploitation likelihood (EPSS) against impact (CVSS). Outside the shaded patch-first corner.
Where this sits among everything scored
Of 382,621 CVEs with a current EPSS score, this one falls in the < 10% band (highlighted). Counts from FIRST.org, log-scaled.
Real-World Exposure
langgraph-apiReal-time download stats are indexed for npm and PyPI packages. This vulnerability affects PyPI packages — download data is not available via public APIs for these ecosystems.
Description
Summary
In affected versions of langgraph-api (the LangGraph Server runtime), a run or cron could be created with a relative webhook target. When the server later delivers such a webhook, it routes the request back into the same application through an in-process loopback transport that the authentication middleware treats as internal and does not authenticate. As a result, a relative webhook target could reach the server's own routes that operate on threads and runs without the authentication context that applies to ordinary external requests.
In deployments that scope threads and runs by owner, this could allow a request associated with one user to reach routes operating on another user's thread, resulting in creation of a run on (or modification of the state of) a thread owned by another user, even where the corresponding direct external requests were correctly denied. Limited metadata from the targeted thread may be incorporated into the created run record.
We have no evidence of this behavior occurring in the wild.
Affected users / systems
You may be affected if you:
- run
langgraph-api(the LangGraph Server / Agent Server runtime, including via the LangGraph Platform Helm chart), - allow runs or crons to specify webhook targets, and
- rely on per-user authorization to separate threads and runs between users.
Impact
- Integrity: creation of a run on, or modification of the state of, a thread owned by another user, beyond the requesting user's authorization scope.
- Confidentiality: limited exposure of another user's thread metadata, incorporated into the created run record.
Patches / mitigation
The webhook URL policy now denies loopback delivery by default: the webhooks.url.disable_loopback policy defaults to enabled. This covers relative webhook targets routed through the in-process transport, as well as localhost-style hostnames, loopback address ranges, and hostnames that resolve into the loopback range. Deployments that legitimately deliver webhooks to a route hosted on the same process can opt back in by setting webhooks.url.disable_loopback: false in langgraph.json (or the equivalent LANGGRAPH_WEBHOOKS configuration); do so only when you control the routes those webhooks reach, as they are delivered without authentication. Fixed in langgraph-api 0.10.0.
Operational guidance
- Upgrade to a release containing this change and keep loopback webhook delivery disabled unless required.
- If loopback delivery is enabled, restrict it to routes you control and apply authorization within those routes.
Affected Packages
| Ecosystem | Package | Vulnerable range | Fix |
|---|---|---|---|
| 🐍PyPI | langgraph-api | all versions | 0.10.0pip install --upgrade 'langgraph-api==0.10.0' |
Detection & mitigation playbook
Open-source dependencyDetect
Scan your dependency tree (package-lock.json, pnpm-lock.yaml, requirements.txt, go.sum, etc.) for langgraph-api, including transitive dependencies — a direct dependency you never call can still pull in a vulnerable version.
Fix
Update langgraph-api to 0.10.0 or later, then make sure no transitive (indirect) dependency still pins the vulnerable range — O3 confirms CVE-2026-55235 is resolved across your whole dependency graph.
Workarounds
Put an independent control in front of the weakness: restrict the affected endpoint or interface to trusted networks, require an additional authentication factor or proxy-level check, and invalidate existing sessions and credentials in case the flaw has already been used.
Frequently Asked Questions
Is CVE-2026-55235 in your dependencies?
Find it across PyPI, including transitive dependencies.