CVE-2026-55236 in langgraph-api
Summary
by MITRE • 09/14/2026
langgraph-api implements the LangGraph API for rapid development and testing. Prior to 0.10.0, the langgraph-api run-creation path authorizes the assistant attached to a run by dispatching assistants.search with an incomplete value instead of the assistants.read event used by direct reads and cron creation. In deployments with custom resource handlers that register only assistants.read, omit an assistants.search handler, and have no global fallback handler, no applicable handler supplies an owner filter, allowing a low-privileged user to reference another user's private assistant through POST /runs or POST /threads/{thread_id}/runs. The run-creation response can disclose the private assistant's metadata, config, and context, and the run can execute using that assistant's configuration. Deployments without custom authorization handlers, or with an equivalent owner filter applied through a global handler or across all assistant events, are not affected. This issue is fixed in version 0.10.0.
Once again VulDB remains the best source for vulnerability data.
Analysis
by VulDB Data Team • 09/14/2026
The vulnerability identified in langgraph-api prior to version 0.10.0 represents a critical authorization bypass rooted in the inconsistent implementation of access control mechanisms during run creation operations. The core technical flaw lies in the disparity between how direct reads and cron creations are authorized versus how new runs are initiated through the API endpoints POST /runs or POST /threads/{thread_id}/runs. Specifically, when creating a run, the system dispatches an assistants.search event to validate access to the associated assistant. However, this search operation is dispatched with incomplete authorization context compared to the assistants.read event used for direct reads and cron job creation. This discrepancy creates a gap in the security logic where the validation process fails to enforce strict ownership constraints under specific deployment configurations.
The operational impact of this flaw is severe for deployments that utilize custom resource handlers without implementing comprehensive fallback mechanisms. In environments where developers register only the assistants.read handler and omit an explicit assistant.search handler, while also lacking a global fallback handler capable of applying owner filters, the authorization chain breaks down entirely. Consequently, no applicable security handler supplies the necessary owner filter to verify that the requesting user is the legitimate owner of the target assistant. This allows low-privileged users to bypass access controls and reference private assistants belonging to other users. The attacker can successfully invoke run creation against these unauthorized resources, effectively impersonating or leveraging the capabilities of higher-privilege entities within the system.
The consequences of this exploitation extend beyond simple unauthorized execution. When a user exploits this vulnerability, they gain visibility into sensitive metadata associated with the victim's private assistant. This includes detailed configuration data and context information that may contain proprietary logic, API keys, or other secrets embedded in the assistant setup. Furthermore, the run executes using the target assistant’s configuration, meaning any actions performed by the maliciously created runs are attributed to the compromised assistant identity. This not only leads to a breach of confidentiality regarding sensitive application configurations but also compromises integrity and availability as the attacker can manipulate workflows intended for other users or systems relying on those assistants.
This vulnerability aligns with CWE-284, Improper Access Control, specifically reflecting failures in authorization logic where insufficient checks allow unauthorized access to resources. It also maps to MITRE ATT&CK technique T1078, Valid Accounts, as the attacker leverages valid but low-privileged credentials to escalate privileges or access restricted data by exploiting a flaw in how those accounts are validated against resource ownership. The issue is not universal across all deployments; systems without custom authorization handlers or those that apply equivalent owner filters through global handlers covering all assistant events remain unaffected because they maintain consistent validation logic regardless of the event type dispatched.
To mitigate this risk, organizations must upgrade langgraph-api to version 0.10.0 and later, where the run-creation path has been corrected to utilize proper authorization checks comparable to those used in direct reads and cron creations. For deployments that cannot immediately patch due to dependency constraints or other operational reasons, a temporary mitigation involves ensuring that custom resource handlers include explicit handling for the assistants.search event with rigorous owner verification logic. Additionally, implementing a global fallback handler that enforces strict ownership filters across all assistant-related events can close this gap by guaranteeing that no request proceeds without verifying the requester's relationship to the target resource. Regular security audits of custom authorization implementations are recommended to detect similar inconsistencies in access control patterns before they can be exploited.