Middleware

Middleware provides a way to more tightly control what happens inside the agent. The core agent loop involves calling a model, letting it choose tools to execute, and then finishing when it calls no more tools.

Middleware provides control over what happens before and after those steps. Each middleware can add in three different types of modifiers:

Middleware.beforeModel: runs before model execution. Can update state or jump to a different node (model, tools, __end__)
Middleware.modifyModelRequest: runs before model execution, to prepare the model request object. Can only modify the current model request object (no permanent state updates) and cannot jump to a different node.
Middleware.afterModel: runs after model execution, before tools are executed. Can update state or jump to a different node (model, tools, __end__)

An agent can contain multiple middleware. Each middleware does not need to implement all three of beforeModel, modifyModelRequest, afterModel.

Using in an agent

import { createAgent } from "langchain";
import { summarizationMiddleware, humanInTheLoopMiddleware } from "langchain/middleware";

const agent = createAgent({
    // ...
    middleware: [summarizationMiddleware, humanInTheLoopMiddleware],
    // ...
});

Middleware is highly flexible and replaces some other functionality in the agent. As such, when middleware are used, there are some restrictions on the arguments used to create the agent:

model must be either a string or a BaseChatModel. Will error if a function is passed. If you want to dynamically control the model, use AgentMiddleware.modifyModelRequest
prompt must be either a string or None. Will error if a function is passed. If you want to dynamically control the prompt, use AgentMiddleware.modifyModelRequest
preModelHook must not be provided. Use AgentMiddleware.beforeModel instead.
postModelHook must not be provided. Use AgentMiddleware.afterModel instead.

Built-in middleware

LangChain provides several built in middleware to use off-the-shelf

Summarization
Human-in-the-loop
Anthropic prompt caching

Summarization

The summarizationMiddleware automatically manages conversation history by summarizing older messages when token limits are approached. This middleware monitors the total token count of messages and creates concise summaries to preserve context while staying within model limits. Key keatures:

Automatic token counting and threshold monitoring
Intelligent message partitioning that preserves AI/Tool message pairs
Customizable summary prompts and token limits

Use Cases:

Long-running conversations that exceed token limits
Multi-turn dialogues with extensive context

import { createAgent } from "langchain";
import { summarizationMiddleware } from "langchain/middleware";
import { ChatOpenAI } from "@langchain/openai";

const agent = createAgent({
    model: "openai:gpt-4o",
    tools: [weatherTool, calculatorTool],
    middleware: [
        summarizationMiddleware({
            model: new ChatOpenAI({ model: "gpt-4o-mini" }),
            maxTokensBeforeSummary: 4000, // Trigger summarization at 4000 tokens
            messagesToKeep: 20, // Keep last 20 messages after summary
            summaryPrompt: "Custom prompt for summarization...", // Optional
        }),
    ],
});

Configuration options:

model: Language model to use for generating summaries (required)
maxTokensBeforeSummary: Token threshold that triggers summarization
messagesToKeep: Number of recent messages to preserve (default: 20)
tokenCounter: Custom function for counting tokens (defaults to character-based approximation)
summaryPrompt: Custom prompt template for summary generation
summaryPrefix: Prefix added to system messages containing summaries (default: ”## Previous conversation summary:”)

The middleware ensures tool call integrity by:

Never splitting AI messages from their corresponding tool responses
Preserving the most recent messages for continuity
Including previous summaries in new summarization cycles

Human-in-the-loop

The humanInTheLoopMiddleware enables human oversight and intervention for tool calls made by AI agents. This middleware intercepts tool executions and allows human operators to approve, modify, reject, or manually respond to tool calls before they execute. Key features:

Selective tool approval based on configuration
Multiple response types (accept, edit, ignore, response)
Asynchronous approval workflow using LangGraph interrupts
Custom approval messages with contextual information

Use cases:

High-stakes operations requiring human approval (database writes, file system changes)
Quality control and safety checks for AI actions
Compliance scenarios requiring audit trails
Development and testing of agent behaviors

import { createAgent } from "langchain";
import { humanInTheLoopMiddleware } from "langchain/middleware";
import { Command } from "@langchain/langgraph";

const agent = createAgent({
    model: "openai:gpt-4o",
    tools: [writeFileTool, executeSQLTool, readDataTool],
    middleware: [
        humanInTheLoopMiddleware({
            toolConfigs: {
                write_file: {
                    requireApproval: true,
                    description: "⚠️ File write operation requires approval",
                },
                execute_sql: {
                    requireApproval: true,
                    description: "🚨 SQL execution requires DBA approval",
                },
                read_data: {
                    requireApproval: false, // Safe operation, no approval needed
                },
            },
            messagePrefix: "Tool execution pending approval",
        }),
    ],
    checkpointer: new MemorySaver(), // Required for interrupts
});

Handling approval requests: When a tool requires approval, the agent execution pauses and waits for human input:

// Initial invocation
const result = await agent.invoke(
    {
        messages: [new HumanMessage("Delete old records from the database")],
    },
    config
);

// Check if paused for approval
const state = await agent.graph.getState(config);
if (state.next?.length > 0) {
    const requests = state.tasks?.[0]?.interrupts?.[0]?.value;

    // Display tool details to human
    console.log("Tool:", requests[0].action);
    console.log("Arguments:", requests[0].args);

    // Resume with approval decision
    await agent.invoke(
        new Command({
            resume: [{ type: "accept" }], // or "edit", "ignore", "response"
        }),
        config
    );
}

Response types:

accept: Execute the tool with original arguments
edit: Modify arguments before execution - { type: "edit", args: { action: "tool_name", args: { modified: "args" } } }
ignore: Skip tool execution and terminate agent
response: Provide manual response instead of executing tool - { type: "response", args: "Manual response text" }

Configuration:

toolConfigs: Map of tool names to their approval settings
- requireApproval: Whether the tool needs human approval
- description: Custom message shown during approval request
messagePrefix: Default prefix for approval messages

The middleware processes tool calls in order, bundling multiple approval requests into a single interrupt for efficiency. Tools not requiring approval execute immediately without interruption.

Anthropic prompt caching

AnthropicPromptCachingMiddleware is a middleware that enables you to enable Anthropic’s native prompt caching. Prompt caching enables optimal API usage by allowing resuming from specific prefixes in your prompts. This is particularly useful for tasks with repetitive prompts or prompts with redundant information.

Learn more about Anthropic Prompt Caching (strategies, limitations, etc.) here.

When using prompt caching, you’ll likely want to use a checkpointer to store conversation history across invocations.

import { createAgent, HumanMessage, MemorySaver } from "langchain";
import { anthropicPromptCachingMiddleware } from "langchain/middleware";

const LONG_PROMPT = `
Please be a helpful assistant.

<Lots more context ...>
`;

const checkpointer = new MemorySaver();
const agent = createAgent({
    model: "anthropic:claude-sonnet-4-latest",
    prompt: LONG_PROMPT,
    middleware: [anthropicPromptCachingMiddleware({ ttl: "5m" })],
    checkpointer,
});
const config = { configurable: { thread_id: "1" } };

const result = await agent.invoke({ messages: [HumanMessage("Hi, my name is Bob")] }, config);

const result = await agent.invoke({ messages: [HumanMessage("What's my name?")] }, config);

Custom Middleware

Middleware for agents are subclasses of AgentMiddleware, which implement one or more of its hooks. AgentMiddleware currently provides three different ways to modify the core agent loop:

before_model: runs before the model is run. Can update state or exit early with a jump.
modify_model_request: runs before the model is run. Cannot update state or exit early with a jump.
after_model: runs after the model is run. Can update state or exit early with a jump.

In order to exit early, you can add a jump_to key to the state update with one of the following values:

"model": Jump to the model node
"tools": Jump to the tools node
"__end__": Jump to the end node

If this is specified, all subsequent middleware will not run. Learn more about exiting early in the agent jumps section.

`before_model`

Runs before the model is run. Can modify state by returning a new state object or state update. Signature:

import type { AgentState, AgentJump } from "langchain";

function beforeModel(state: AgentState): AgentState | AgentJump | undefined {
    // ...
}

`modify_model_request`

Runs before the model has run, but after all the before_model calls. These functions cannot modify permanent state or exit early. Rather, they are intended to modify calls to the model in a **stateless* way. If you want to modify calls to the model in a **stateful** way, you will need to use before_model Modifies the model request. The model request has several key properties:

model (BaseChatModel): the model to use. Note: this needs to the base chat model, not a string.
system_prompt (str): the system prompt to use. Will get prepended to messages
messages (list of messages): the message list. Should not include system prompt.
tool_choice (Any): the tool choice to use
tools (list of BaseTool): the tools to use for this model call
responseFormat (ResponseFormat): the response format to use for structured output

Signature:

function modifyModelRequest (request: ModelRequest, state: AgentState): ModelRequest {
    return request;
}

`after_model`

Runs after the model is run. Can modify state by returning a new state object or state update. Signature:

import type { AgentState, AgentUpdate, AgentJump } from "langchain";

function afterModel (state: AgentState): AgentUpdate | AgentJump | undefined {
    // ...
}

New state keys

Middleware can extend the agent’s state with custom properties, enabling rich data flow between middleware components and ensuring type safety throughout the agent execution.

State extension

Middleware can define additional state properties that persist throughout the agent’s execution. These properties become part of the agent’s state and are available to all hooks for said middleware. When a middleware defines required state properties through its stateSchema, these properties must be provided when invoking the agent:

import { createMiddleware, createAgent } from "langchain";
import { z } from "zod";

// Middleware with custom state requirements
const authMiddleware = createMiddleware({
    name: "AuthMiddleware",
    stateSchema: z.object({
        userId: z.string(), // Required
        userRole: z.string().default("user"), // Optional with default
    }),
    beforeModel: (state) => {
        // Access custom state properties
        console.log(`User ${state.userId} with role ${state.userRole}`);
        return;
    },
});

const agent = createAgent({
    model: "openai:gpt-4o",
    tools: [],
    middlewares: [authMiddleware] as const,
});

// TypeScript enforces required state properties
const result = await agent.invoke({
    messages: [new HumanMessage("Hello")],
    userId: "user-123", // Required by middleware
    // userRole is optional due to default value
});

Context extension

This is currently only available in JavaScript.

Context properties are configuration values passed through the runnable config. Unlike state, context is read-only and typically used for configuration that doesn’t change during execution. Middleware can define context requirements that must be satisfied through the agent’s configuration:

const rateLimitMiddleware = createMiddleware({
    name: "RateLimitMiddleware",
    contextSchema: z.object({
        maxRequestsPerMinute: z.number(),
        apiKey: z.string(),
    }),
    beforeModel: async (state, runtime) => {
        // Access context through runtime
        const { maxRequestsPerMinute, apiKey } = runtime.context;

        // Implement rate limiting logic
        const allowed = await checkRateLimit(apiKey, maxRequestsPerMinute);
        if (!allowed) {
            return { jump_to: "END" };
        }
    },
});

// Context is provided through config
await agent.invoke(
    { messages: [new HumanMessage("Process data")] },
    {
        context: {
            maxRequestsPerMinute: 60,
            apiKey: "api-key-123",
        },
    }
);

Combining multiple middleware

When using multiple middleware, their state and context schemas are merged. All required properties from all middleware must be satisfied:

const middleware1 = createMiddleware({
    name: "Middleware1",
    stateSchema: z.object({
        prop1: z.string(),
        sharedProp: z.number(),
    }),
});

const middleware2 = createMiddleware({
    name: "Middleware2",
    stateSchema: z.object({
        prop2: z.boolean(),
        sharedProp: z.number(), // Same property name must have compatible types
    }),
});

const agent = createAgent({
    model: "openai:gpt-4o",
    tools: [],
    middlewares: [middleware1, middleware2] as const,
});

// Must provide all required properties
const result = await agent.invoke({
    messages: [new HumanMessage("Hello")],
    prop1: "value1", // Required by middleware1
    prop2: true, // Required by middleware2
    sharedProp: 42, // Required by both
});

Agent-level context schema

Agents can also define their own context requirements that combine with middleware requirements:

const agent = createAgent({
    model: "openai:gpt-4o",
    tools: [],
    contextSchema: z.object({
        environment: z.enum(["development", "production"]),
    }),
    middlewares: [rateLimitMiddleware] as const,
});

// Must satisfy both agent and middleware context requirements
await agent.invoke(
    { messages: [new HumanMessage("Deploy application")] },
    {
        context: {
            environment: "production", // Required by agent
            maxRequestsPerMinute: 60, // Required by middleware
            apiKey: "api-key-123", // Required by middleware
        },
    }
);

Best practices

Use State for Dynamic Data: Properties that change during execution (user session, accumulated data)
Use Context for Configuration: Static configuration values (API keys, feature flags, limits)
Provide Defaults When Possible: Use .default() in Zod schemas to make properties optional
Document Requirements: Clearly document what state and context properties your middleware requires
Type Safety: Leverage TypeScript’s type checking to catch missing properties at compile time

The type system ensures all required properties are provided, preventing runtime errors:

// TypeScript error: Property 'userId' is missing
await agent.invoke({
    messages: [new HumanMessage("Hello")],
    // userId is required but not provided
});

// TypeScript error: Type 'number' is not assignable to type 'string'
await agent.invoke({
    messages: [new HumanMessage("Hello")],
    userId: 123, // Wrong type
});

Middleware execution order

You can provide multiple middlewares. They are executed in the following logic: beforeModel: Are run in the order they are passed in. If an earlier middleware exits early, then following middleware are not run modifyModelRequest: Are run in the order they are passed in. afterModel: Are run in the reverse order that they are passed in. If an earlier middleware exits early, then following middleware are not run

Agent jumps

In order to exit early, you can add a jump_to key to the state update with one of the following values:

"model": Jump to the model node
"tools": Jump to the tools node
"__end__": Jump to the end node

If this is specified, all subsequent middleware will not run. If you jump to model node, all beforeModel middleware will run. It’s forbidden to jump to model from an existing beforeModel middleware. Example usage:

import type { AgentState, AgentJump } from "langchain";

afterModel (state: AgentState): AgentState | AgentJump | undefined {
    return {
        messages: ...,
        jump_to: "model"
    }
}

Get started

Core components

Advanced usage

Use in production

Using in an agent

Built-in middleware

Summarization

Human-in-the-loop

Anthropic prompt caching

Custom Middleware

`before_model`

`modify_model_request`

`after_model`

New state keys

State extension

Context extension

Combining multiple middleware

Agent-level context schema

Best practices

Middleware execution order

Agent jumps

Get started

Core components

Advanced usage

Use in production

​Using in an agent

​Built-in middleware

​Summarization

​Human-in-the-loop

​Anthropic prompt caching

​Custom Middleware

​before_model

​modify_model_request

​after_model

​New state keys

​State extension

​Context extension

​Combining multiple middleware

​Agent-level context schema

​Best practices

​Middleware execution order

​Agent jumps

Using in an agent

Built-in middleware

Summarization

Human-in-the-loop

Anthropic prompt caching

Custom Middleware

`before_model`

`modify_model_request`

`after_model`

New state keys

State extension

Context extension

Combining multiple middleware

Agent-level context schema

Best practices

Middleware execution order

Agent jumps