AI Sec Watch: A Security Intelligence Platform for AI Systems

Luu, T.J.

GHSA-2c9q-c2q9-qgqv: langgraph-api: Relative webhook targets in LangGraph Server can reach in-process routes without authentication

mediumvulnerability

security

Source: GitHub Advisory DatabaseAugust 19, 2026CVE-2026-55235

Summary

LangGraph Server had a security flaw where webhooks (automated messages sent to other services) with relative targets could bypass authentication checks by routing requests internally without verifying the user's identity. This could allow one user to create or modify runs (execution records) on threads (conversation sessions) owned by another user, and view limited information about other users' threads.

Solution / Mitigation

Upgrade to langgraph-api version 0.10.0 or later. The webhook URL policy now has `webhooks.url.disable_loopback` enabled by default, which blocks loopback delivery (requests sent back to the same server). If your deployment legitimately needs to send webhooks to routes on the same process, you can set `webhooks.url.disable_loopback: false` in `langgraph.json` or the equivalent `LANGGRAPH_WEBHOOKS` environment variable configuration, but only if you control those routes and apply authorization checks within them.