Langchain Middleware

langchain-ai/skills-benchmarks-langchain-middleware

INVOKE THIS SKILL when you need human-in-the-loop approval, custom middleware, or structured output. Covers HumanInTheLoopMiddleware for human approval of dangerous tool calls, creating custom middleware with hooks, Command resume patterns, and structured output with Pydantic/Zod.

3k tokens

context cost

the whole folder, loaded on every use

files

instructions only

copies elsewhere

how many repositories repackaged it

111

stars on the repo

on the repository, not the skill itself

Install

one command, takes just this skill from the repository

npx skills add https://github.com/langchain-ai/skills-benchmarks --skill langchain-middleware

The instruction itself

4 sections, as written by the author

Middleware patterns for production LangChain agents:

HumanInTheLoopMiddleware / humanInTheLoopMiddleware: Pause before dangerous tool calls for human approval
Custom middleware: Intercept tool calls for error handling, logging, retry logic
Command resume: Continue execution after human decisions (approve, edit, reject)

Requirements: Checkpointer + thread_id config for all HITL workflows.

</overview>

Human-in-the-Loop

<ex-basic-hitl-setup>

Set up an agent with HITL middleware that pauses before sending emails for approval.

from langchain.agents import create_agent
from langchain.agents.middleware import HumanInTheLoopMiddleware
from langgraph.checkpoint.memory import MemorySaver
from langchain.tools import tool

@tool
def send_email(to: str, subject: str, body: str) -> str:
    """Send an email."""
    return f"Email sent to {to}"

agent = create_agent(
    model="gpt-4.1",
    tools=[send_email],
    checkpointer=MemorySaver(),  # Required for HITL
    middleware=[
        HumanInTheLoopMiddleware(
            interrupt_on={
                "send_email": {"allowed_decisions": ["approve", "edit", "reject"]},
            }
        )
    ],
)

</python>

Set up an agent with HITL that pauses before sending emails for human approval.

import { createAgent, humanInTheLoopMiddleware } from "langchain";
import { MemorySaver } from "@langchain/langgraph";
import { tool } from "@langchain/core/tools";
import { z } from "zod";

const sendEmail = tool(
  async ({ to, subject, body }) => `Email sent to ${to}`,
  {
    name: "send_email",
    description: "Send an email",
    schema: z.object({ to: z.string(), subject: z.string(), body: z.string() }),
  }
);

const agent = createAgent({
  model: "anthropic:claude-sonnet-4-5",
  tools: [sendEmail],
  checkpointer: new MemorySaver(),
  middleware: [
    humanInTheLoopMiddleware({
      interruptOn: { send_email: { allowedDecisions: ["approve", "edit", "reject"] } },
    }),
  ],
});

</typescript>

</ex-basic-hitl-setup>

<ex-running-with-interrupts>

Run the agent, detect an interrupt, then resume execution after human approval.

from langgraph.types import Command

config = {"configurable": {"thread_id": "session-1"}}

# Step 1: Agent runs until it needs to call tool
result1 = agent.invoke({
    "messages": [{"role": "user", "content": "Send email to [email protected]"}]
}, config=config)

# Check for interrupt
if "__interrupt__" in result1:
    print(f"Waiting for approval: {result1['__interrupt__']}")

# Step 2: Human approves
result2 = agent.invoke(
    Command(resume={"decisions": [{"type": "approve"}]}),
    config=config
)

</python>

Run the agent, detect an interrupt, then resume execution after human approval.

import { Command } from "@langchain/langgraph";

const config = { configurable: { thread_id: "session-1" } };

// Step 1: Agent runs until it needs to call tool
const result1 = await agent.invoke({
  messages: [{ role: "user", content: "Send email to [email protected]" }]
}, config);

// Check for interrupt
if (result1.__interrupt__) {
  console.log(`Waiting for approval: ${result1.__interrupt__}`);
}

// Step 2: Human approves
const result2 = await agent.invoke(
  new Command({ resume: { decisions: [{ type: "approve" }] } }),
  config
);

</typescript>

</ex-running-with-interrupts>

<ex-editing-tool-arguments>

Edit the tool arguments before approving when the original values need correction.

# Human edits the arguments — edited_action must include name + args
result2 = agent.invoke(
    Command(resume={
        "decisions": [{
            "type": "edit",
            "edited_action": {
                "name": "send_email",
                "args": {
                    "to": "[email protected]",  # Fixed email
                    "subject": "Project Meeting - Updated",
                    "body": "...",
                },
            },
        }]
    }),
    config=config
)

</python>

Edit the tool arguments before approving when the original values need correction.

// Human edits the arguments — editedAction must include name + args
const result2 = await agent.invoke(
  new Command({
    resume: {
      decisions: [{
        type: "edit",
        editedAction: {
          name: "send_email",
          args: {
            to: "[email protected]",  // Fixed email
            subject: "Project Meeting - Updated",
            body: "...",
          },
        },
      }]
    }
  }),
  config
);

</typescript>

</ex-editing-tool-arguments>

<ex-rejecting-with-feedback>

Reject a tool call and provide feedback explaining why it was rejected.

# Human rejects
result2 = agent.invoke(
    Command(resume={
        "decisions": [{
            "type": "reject",
            "feedback": "Cannot delete customer data without manager approval",
        }]
    }),
    config=config
)

</python>

</ex-rejecting-with-feedback>

<ex-multiple-tools-different-policies>

Configure different HITL policies for each tool based on risk level.

agent = create_agent(
    model="gpt-4.1",
    tools=[send_email, read_email, delete_email],
    checkpointer=MemorySaver(),
    middleware=[
        HumanInTheLoopMiddleware(
            interrupt_on={
                "send_email": {"allowed_decisions": ["approve", "edit", "reject"]},
                "delete_email": {"allowed_decisions": ["approve", "reject"]},  # No edit
                "read_email": False,  # No HITL for reading
            }
        )
    ],
)

</python>

</ex-multiple-tools-different-policies>

What You CAN Configure

Which tools require approval (per-tool policies)
Allowed decisions per tool (approve, edit, reject)
Custom middleware hooks: before_model, after_model, wrap_tool_call, before_agent, after_agent
Tool-specific middleware (apply only to certain tools)

</boundaries>

Custom Middleware Hooks

Six decorator hooks are available. Two patterns:

Wrap hooks (wrap_tool_call, wrap_model_call): (request, handler) — call handler(request) to proceed, or return early to short-circuit.
Before/after hooks (before_model, after_model, before_agent, after_agent): (state, runtime) — inspect or modify state. Return None or a dict of state updates.

<ex-wrap-tool-call>

@wrap_tool_call intercepts tool execution. Do NOT use yield — it creates a generator and causes NotImplementedError.

from langchain.agents.middleware import wrap_tool_call

@wrap_tool_call
def retry_middleware(request, handler):
    for attempt in range(3):
        try:
            return handler(request)
        except Exception:
            if attempt == 2:
                raise

@wrap_tool_call
def guard_middleware(request, handler):
    if request.tool_call["name"] == "dangerous_tool":
        return "This tool is disabled"  # short-circuit
    return handler(request)

</python>

createMiddleware({ wrapToolCall }) intercepts tool execution.

import { createMiddleware } from "langchain";

const retryMiddleware = createMiddleware({
  wrapToolCall: async (request, handler) => {
    for (let attempt = 0; attempt < 3; attempt++) {
      try { return await handler(request); }
      catch (e) { if (attempt === 2) throw e; }
    }
  },
});

</typescript>

</ex-wrap-tool-call>

<ex-before-after-hooks>

before_model / after_model / before_agent / after_agent all share (state, runtime) signature.

from langchain.agents.middleware import before_model, after_model

@before_model
def log_calls(state, runtime):
    print(f"Calling model with {len(state['messages'])} messages")

@after_model
def check_output(state, runtime):
    print(f"Model responded")

</python>

All before/after hooks share the same (state, runtime) signature via createMiddleware.

import { createMiddleware } from "langchain";

const loggingMiddleware = createMiddleware({
  beforeModel: (state, runtime) => {
    console.log(`Calling model with ${state.messages.length} messages`);
  },
  afterModel: (state, runtime) => {
    console.log("Model responded");
  },
});

</typescript>

</ex-before-after-hooks>

What You CANNOT Configure

Interrupt after tool execution (must be before)
Skip checkpointer requirement for HITL

</boundaries>

<fix-missing-checkpointer>

HITL middleware requires a checkpointer to persist state.

# WRONG
agent = create_agent(model="gpt-4.1", tools=[send_email], middleware=[HumanInTheLoopMiddleware({...})])

# CORRECT
agent = create_agent(
    model="gpt-4.1", tools=[send_email],
    checkpointer=MemorySaver(),  # Required
    middleware=[HumanInTheLoopMiddleware({...})]
)

</python>

HITL requires a checkpointer to persist state.

// WRONG: No checkpointer
const agent = createAgent({
  model: "anthropic:claude-sonnet-4-5", tools: [sendEmail],
  middleware: [humanInTheLoopMiddleware({ interruptOn: { send_email: true } })],
});

// CORRECT: Add checkpointer
const agent = createAgent({
  model: "anthropic:claude-sonnet-4-5", tools: [sendEmail],
  checkpointer: new MemorySaver(),
  middleware: [humanInTheLoopMiddleware({ interruptOn: { send_email: true } })],
});

</typescript>

</fix-missing-checkpointer>

<fix-no-thread-id>

Always provide thread_id when using HITL to track conversation state.

# WRONG
agent.invoke(input)  # No config!

# CORRECT
agent.invoke(input, config={"configurable": {"thread_id": "user-123"}})

</python>

</fix-no-thread-id>

<fix-wrong-resume-syntax>

Use Command class to resume execution after an interrupt.

# WRONG
agent.invoke({"resume": {"decisions": [...]}})

# CORRECT
from langgraph.types import Command
agent.invoke(Command(resume={"decisions": [{"type": "approve"}]}), config=config)

</python>

Use Command class to resume execution after an interrupt.

// WRONG
await agent.invoke({ resume: { decisions: [...] } });

// CORRECT
import { Command } from "@langchain/langgraph";
await agent.invoke(new Command({ resume: { decisions: [{ type: "approve" }] } }), config);

</typescript>

</fix-wrong-resume-syntax>

How to use it

Copy the folder

Take langchain-ai/skills-benchmarks-langchain-middleware from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.