TypeScript framework for building LLM-powered applications with agents, tools, middleware, and model interoperability
—
Pending
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Pending
The risk profile of this skill
Complete API reference for all LangChain components. Each section provides detailed type definitions, parameter descriptions, and usage examples.
Want task-based examples instead? See the Task Index for "I want to..." → code mappings.
Complete API for creating and executing ReAct agents.
Key APIs:
createAgent() - Create production-ready agentsReactAgent class - Agent execution methods (invoke, stream, batch)API for creating tools that agents can use.
Key APIs:
tool() - Create structured toolsDynamicTool - String-input toolsStructuredTool - Schema-based toolsDynamicStructuredTool - Dynamic structured toolsUniversal model initialization API.
Key APIs:
initChatModel() - Initialize chat models from any providerMessage types and utilities for agent communication.
Key APIs:
HumanMessage - User messagesAIMessage - Assistant messagesSystemMessage - System instructionsToolMessage - Tool resultsfilterMessages() - Filter message arraystrimMessages() - Trim messages to fit contextType-safe structured response generation.
Key APIs:
toolStrategy() - Tool-based structured outputproviderStrategy() - Provider-native structured outputToolStrategy classProviderStrategy classView Structured Outputs API Reference →
Storage implementations for persisting data.
Key APIs:
InMemoryStore - In-memory key-value storeExecution context utilities.
Key APIs:
context.get() - Get context valuescontext.set() - Set context valuescontext.run() - Run with contextcontext.runAsync() - Async context executionDocument representation for text processing.
Key APIs:
Document class - Document containerView Documents API Reference →
// Import commonly used APIs
import {
createAgent,
tool,
initChatModel,
HumanMessage,
AIMessage,
SystemMessage,
toolStrategy,
providerStrategy,
createMiddleware,
InMemoryStore,
} from "langchain";function createAgent<TConfig>(params: CreateAgentParams): ReactAgent<TConfig>;function tool<T = any>(
func: (input: T, config?: ToolConfig) => any | Promise<any>,
fields: {
name: string;
description: string;
schema: ZodType<T>;
}
): StructuredTool<T>;function initChatModel<RunInput = any, CallOptions extends BaseChatModelCallOptions = BaseChatModelCallOptions>(
model?: string | ChatModel,
fields?: InitChatModelFields
): ChatModel<RunInput, CallOptions>;// Infer types from agents
type InferAgentState<T>
type InferAgentResponse<T>
type InferAgentContext<T>
type InferAgentTools<T>
type InferAgentMiddleware<T>
// Infer types from middleware
type InferMiddlewareState<T>
type InferMiddlewareStates<T extends readonly AgentMiddleware[]>
type InferMiddlewareContext<T>
type InferMiddlewareContexts<T extends readonly AgentMiddleware[]>
type InferMiddlewareTools<T extends readonly AgentMiddleware[]>
// Schema inference
type InferSchemaInput<A>
type InferContextInput<ContextSchema>Quick lookup for all error classes:
MultipleToolsBoundError - Tools already bound to modelMultipleStructuredOutputsError - Multiple outputs returnedStructuredOutputParsingError - Failed to parse outputToolInvocationError - Tool execution failedToolCallLimitExceededError - Tool call limit exceededPIIDetectionError - PII detection failedComplete error handling guide →
// Agent types
type CreateAgentParams
type ReactAgent<TConfig>
type UserInput<TStateSchema>
type State<TConfig>
type InvokeConfiguration<ContextSchema>
type StreamConfiguration<ContextSchema>
type BatchConfiguration<ContextSchema>
// Tool types
type Tool
type ToolCall
type ToolResult
type ToolConfig
type ToolRuntime
// Message types
type BaseMessage
type MessageInput
type ContentBlock
// Middleware types
type AgentMiddleware<TSchema, TContextSchema, TFullContext, TTools>
type MiddlewareConfig<TSchema, TContextSchema, TTools>
// Response format types
type ResponseFormat
type JsonSchemaFormat
// Runtime types
type Runtime<TContext>Getting Started
API References
Advanced Topics
Middleware
Integrations
Building Agents
Adding Tools
Structured Outputs
Using Middleware
Type Safety
docs