LangGraph Human-in-the-Loop Skill for Claude Code

Teaches Claude the right way to pause a LangGraph agent for human approval and resume it, in Python and TypeScript.

Installs: 17.6K; on skills.sh
Stars: 1.3K; on GitHub
Updated: Oct 2026; last repo commit

This is a reference skill with no scripts. It gives the assistant the LangGraph patterns for pausing a run with interrupt(), showing a value to the user, and continuing with Command(resume=...). It covers a basic pause, an approval workflow, a validation loop, several interrupts at once, and side effects that must be safe to repeat. Each pattern has Python and TypeScript code.

Its description tells the agent to load it when you are implementing human-in-the-loop, pausing for approval, or handling errors in LangGraph. The most useful warning: when a graph resumes, the node restarts from its beginning, so code before interrupt() runs again. It also says a checkpointer and a thread ID are required.

Good for

Developers adding approval steps or input checks to a LangGraph agent
Teams whose paused graphs create duplicate records or seem stuck after resume
Python or TypeScript users of LangGraph with a coding assistant

Install langgraph-human-in-the-loop

Claude Code

Copy this prompt.

Install the langgraph-human-in-the-loop skill in Claude Code. Run this command:

npx skills add https://github.com/langchain-ai/langchain-skills --skill langgraph-human-in-the-loop

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

From its listing on skills.sh. Source.

What's in the skill

When it loads

INVOKE THIS SKILL when implementing human-in-the-loop patterns, pausing for approval, or handling errors in LangGraph. Covers interrupt(), Command(resume=...), approval/validation workflows, and the 4-tier error handling strategy.

Sections in SKILL.md

Requirements
Basic Interrupt + Resume
Approval Workflow
Validation Loop
Multiple Interrupts
Side Effects Before Interrupt Must Be Idempotent
Subgraph re-execution on resume
Command(resume) Warning
Fixes
What You Should NOT Do

Key facts

Author: langchain-ai
Repo: langchain-ai/langchain-skills
License: MIT (as listed by the crawl)
Installs on skills.sh: 17,596 (Nov 2, 2026)
GitHub stars: 1,320 (repo, Nov 2, 2026)
Last commit: 2026-10-29
Files: 1 file: SKILL.md
Covers: LangGraph in Python and TypeScript
Status: The repo README says the project is in early development

What it includes

A checkpointer, a thread ID and a JSON-serializable payload are all required
InMemorySaver for development, PostgresSaver for production. Every invoke or stream call needs the same thread_id.
Requirements
interrupt(value) pauses; Command(resume=value) continues with that value
The paused value appears under __interrupt__ in the result. The resume value becomes the return value of interrupt().
Basic interrupt and resume
Show a draft, then route on the human's decision
The pattern for approve or reject steps before an agent acts.
Approval workflow
Re-ask until the human input passes a check
The example asks for an age and asks again when the answer is not valid.
Validation loop
Parallel nodes can pause at once; resume them together with a map of id to value
Each pending interrupt has an ID, and you answer all of them in one resume.
Multiple interrupts
Anything before interrupt() runs again on resume, so use upserts or move it after
Inserts and list appends before an interrupt create duplicates every time the graph resumes. Subgraphs re-run both the parent node and the subgraph node.
Idempotent side effects
Only Command(resume=...) should be passed as invoke input, never Command(update=...)
Passing update as input resumes from the last checkpoint and the graph looks stuck.
Command(resume) warning
Wrong and correct code for a missing checkpointer, a changed thread ID and resume syntax
Short before-and-after snippets in Python and TypeScript.
Fixes

From the official docs

Questions

What does the LangGraph human-in-the-loop skill do?

It gives the assistant correct, working patterns for pausing a LangGraph run and resuming it with human input. It covers approvals, input validation, parallel interrupts and the traps that cause duplicate records or a stuck graph.

How do I install it?

Use the install box above for this one skill. The README also shows installing every skill in the repo with npx skills add langchain-ai/langchain-skills --skill '*' --yes, and a Claude Code plugin route.

Does it work in Claude Code, Cursor and Codex?

The README says the skills install through the skills CLI for any agent that supports the Agent Skills spec, and names Claude Code, Cursor and Windsurf. It gives install commands for Claude Code, Codex and GitHub Copilot.

Do I need an API key?

No key is mentioned in SKILL.md. The skill is plain instructions and code samples with no bundled files. Your own LangGraph project will need whatever model keys it uses.

Why does my code before interrupt() run twice?

When the graph resumes, the node restarts from its beginning, so everything before interrupt() runs again. Use upserts or check-before-create, or put the side effect after the interrupt or in its own node.

Why does my graph look stuck after I resume?

SKILL.md lists passing Command(update=...) as invoke input as a cause. Resume with Command(resume=...) and the same thread_id instead.

Does it cover error handling?

Only one tier of it. Errors the user can fix are handled by pausing with interrupt(). For the full four-tier strategy the skill points to the fundamentals skill, which the README lists as langgraph-fundamentals.

Is it safe?

It is a single instruction file with code examples and bundles no scripts. Our advice: review the code it makes the assistant write, especially anything that acts after an approval.

Information

updated Nov 2, 2026

Repository: langchain-ai/langchain-skills
Folder: config/skills/langgraph-human-in-the-loop

langchain-ai/langchain-skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 4

langgraph-human-in-the-loop SKILL.md · github.com
langchain-skills README · github.com
skills.sh listing · skills.sh
langchain-ai/langchain-skills · github.com