Overview

Relevant source files

Purpose and Scope

Langflow is an open-source platform for building and deploying AI-powered agents and workflows through visual authoring and programmatic APIs. This page provides a high-level overview of the system architecture, package structure, and installation methods. For detailed information about specific subsystems, refer to the child pages linked in this document.

Sources: README.md16-27 pyproject.toml1-21

What is Langflow

Langflow is a multi-layered application for building AI workflows with three core layers:

Frontend: React 19 + TypeScript SPA using @xyflow/react for visual flow editing src/frontend/package.json66-97
Backend: FastAPI server with REST/WebSocket APIs, providing execution and management services src/backend/base/pyproject.toml21
Component System: Dynamic component architecture enabling natural language generation of components via Langflow Assistant README.md88-90 and source code customization README.md21

Workflows can be executed via REST API, deployed as MCP (Model Context Protocol) servers src/lfx/pyproject.toml96 or integrated into various AI frameworks.

Version: 1.11.0 pyproject.toml3 Python Requirements: >=3.10,<3.15 pyproject.toml5 Default Port: 7860 README.md59 Makefile18

Sources: README.md16-27 pyproject.toml2-5 src/frontend/package.json3-97 src/lfx/pyproject.toml96

System Architecture Overview

Diagram: System Architecture - Client Layer to Storage

The FastAPI application initializes core services during the lifespan context. Services are managed through a central service manager pattern, providing access to the database via DatabaseService and configuration via SettingsService. The execution engine translates JSON graph definitions into executable Directed Acyclic Graphs (DAGs).

For details, see System Architecture.

Sources: src/backend/base/pyproject.toml21-48 src/lfx/src/lfx/services/settings/base.py89-104

Three-Tier Package Structure

Langflow uses a monorepo workspace to separate concerns, managed by uv pyproject.toml132-160

Diagram: Package Dependency Hierarchy

langflow pyproject.toml2: Top-level package providing the CLI entry point.
langflow-base src/backend/base/pyproject.toml2: Core package containing the application logic and service layer. It depends on lfx src/backend/base/pyproject.toml20
lfx src/lfx/pyproject.toml2: Lightweight executor and CLI tool for serving flows, integrating with the Model Context Protocol src/lfx/pyproject.toml64-95
langflow-sdk src/sdk/pyproject.toml2: Python client for interacting with the Langflow REST API src/lfx/pyproject.toml60

For details, see Package Structure.

Sources: pyproject.toml132-160 src/backend/base/pyproject.toml1-20 src/lfx/pyproject.toml1-96 src/sdk/pyproject.toml1-4

Frontend Architecture

Diagram: Frontend Stack - React Components to Backend API

The frontend is a React 19 SPA src/frontend/package.json97 using @xyflow/react for the flow canvas src/frontend/package.json66 State management is handled by Zustand src/frontend/package.json129 and the UI is built with Tailwind CSS and Radix UI primitives src/frontend/package.json43-203

Sources: src/frontend/package.json1-129 src/frontend/package.json203

Backend Services

The backend is organized around a service manager pattern. Services are initialized based on environment configuration and the application lifespan.

Service	Purpose	Source
SettingsService	Configuration management via `dynaconf` and `pydantic-settings`	src/backend/base/pyproject.toml41-108
DatabaseService	Persistence using `sqlmodel` and `alembic` migrations	src/backend/base/pyproject.toml39-48
TelemetryService	Telemetry and tracking using OpenTelemetry and Segment	src/backend/base/pyproject.toml67-74
TaskService	Background execution with `anyio` or `celery` backends	src/backend/base/pyproject.toml62
McpComposerService	Pluggable service for per-project MCP orchestration	src/lfx/src/lfx/services/settings/base.py98

For details, see System Architecture.

Sources: src/backend/base/pyproject.toml39-112 src/lfx/src/lfx/services/settings/base.py89-102

Installation and Deployment

Langflow supports multiple installation and deployment methods:

Langflow Desktop: Native installer for Windows and macOS README.md29-34
Docker: Containerized deployment using docker run -p 7860:7860 langflowai/langflow:latest README.md72-78
Python Package (PyPI): Installation via uv pip install langflow README.md46
Source: Development setup using make run_cli README.md68

For details, see Installation and Getting Started.

Sources: README.md29-78

Key Capabilities

Visual Builder: Rapid iteration of AI workflows README.md20
Langflow Assistant: Create components and full workflows using natural language README.md88-90
Extension Bundles: Standalone pip packages for specific providers (e.g., lfx-openai, lfx-anthropic) shipped in lfx-bundles docs/docs/Support/release-notes.mdx107-122 pyproject.toml30-52
MCP Integration: Connect as an MCP client or serve flows as MCP tools README.md25 src/lfx/pyproject.toml96
Multi-agent Orchestration: Complex task execution and conversation management README.md23
Observability: Native integration with LangSmith, LangFuse, and OpenTelemetry README.md26 src/backend/base/pyproject.toml67-74

Sources: README.md18-28 src/backend/base/pyproject.toml67-74 docs/docs/Support/release-notes.mdx107-122

Next Steps

Architecture Details: See System Architecture for component interactions.
Installation: See Installation and Getting Started for setup instructions.
Package Details: See Package Structure for dependency relationships.
Build System: See Build System and Development for development workflows.
Component System: See Component System for component lifecycle and loading.
Backend: See Backend Architecture for service details and API routes.
Frontend: See Frontend Architecture for UI components and state management.