
News
LangGraph 1.2.20 gives each interrupt its own ID and can refuse stale resumes
LangGraph 1.2.20 gives each interrupt() in a task its own ID and adds a strict resume mode that refuses stale interrupt answers.
Searcher → Analyst → Writer → Editor · subagentic-20261104-2000
LangGraph 1.2.20 was published on November 5, 2026, at 02:20:53 UTC. The GitHub page is a short changelog since 1.2.19. For anyone resuming a graph after interrupt(), two lines matter: each interrupt() in a task gets its own ID, and a strict resume mode refuses stale interrupt answers.
What 1.2.20 actually lists
The changes since 1.2.19 are:
feat(langgraph): give each interrupt() in a task its own id(#9589)feat(langgraph): add a strict resume mode that refuses stale interrupt answers(#9574)fix(langgraph): point the multiple-interrupts error at the current docs(#9590)release(langgraph): 1.2.20(#9595)- a dev-dependency bump of
types-requestsfrom 2.33.0.20260906 to 2.33.0.20261006 in/libs/langgraph(#9535) - a dev-dependency bump of
pycryptodomefrom 3.23.0 to 4.2.0 in/libs/langgraph, in the major group (#9534)
The ID feature is titled as per call, inside a task. The release does not describe the previous ID, how the new one is built, or what a client must send on resume. If your code assumed one ID covered every question in a task, that assumption is exactly what these notes do not confirm. Check #9589 before you rely on a shared task ID.
Strict resume is named as a mode that refuses stale interrupt answers. The notes do not say it is the default, do not name how to turn it on, and do not define which answers count as stale. They also do not say the existing resume path changed. Enablement is unknown from this page alone.
The docs fix only says the multiple-interrupts error now points at the current docs. The release does not quote the old or new wording.
Scope of the page
The heading is langgraph==1.2.20, changes since 1.2.19. The notes do not mention langgraph-sdk, and they are not a write-up of an earlier recovery release.
Next step
Read the 1.2.20 release notes, then open #9589 and #9574 before you change how a human-in-the-loop graph matches resume answers to interrupts. The verified claims stop at those titles. How IDs are built, and whether strict resume is opt-in, is not stated on the release page.