langflow

Files (4)

• SKILL.md
• README.md · references/
• file_structure.md · references/
• openrag.md · references/

SKILL.md (first lines)

# Langflow - Visual AI Workflow Platform

## Overview

**Langflow** is an open-source, Python-based platform for building and deploying AI-powered agents and workflows through a visual drag-and-drop interface. With 142,000+ GitHub stars and used by 1,500+ projects, Langflow enables rapid prototyping of AI applications without requiring extensive coding knowledge.

## Key Features

### 🎨 Visual Development Environment
- **Drag-and-Drop Canvas**: Build complex AI workflows visually by connecting component nodes
- **Interactive Playground**: Test and debug flows in real-time without full stack development
- **Component Library**: 200+ pre-built components for LLMs, data sources, agents, tools, and MCP servers
- **Custom Components**: Full Python customization for specialized use cases

### 🤖 Agent & MCP Integration
- **AI Agents**: LLM-powered agents with autonomous tool selection and execution
- **MCP Client & Server**: Built-in Model Context Protocol support for tool integration
- **Multi-Agent Coordination**: Orchestrate multiple agents working together
- **Tool Ecosystem**: Calculator, Web Search, URL fetcher, file operations, and custom tools

### 🚀 Deployment Options
- **Local Development**: Desktop app for macOS/Windows or Python package installation
- **Docker Containers**: Portable, reproducible deployments
- **Kubernetes**: Production-grade orchestration with high availability
- **Cloud Platforms**: Native support for Google Cloud, Hugging Face Spaces, and more
- **API Export**: RESTful API endpoints for external integrations

### 🔧 Production Features
- **Observability**: Integrated LangSmith and LangFuse monitoring
- **Authentication**: API key management with role-based access control
- **Security**: HTTPS support, CORS configuration, reverse proxy compatibility
- **Versioning**: Component version control and flow history
- **JSON Export**: Portable flow definitions for backup and sharing

## Architecture

### Component-Based Design
```
Flow (Workflow)
├── Components (Building Blocks)
│   ├── Inputs/Outputs (Ports)
│   ├── Parameters (Configuration)
│   └── Python Code (Logic)
├── Edges (Connections)
└── Canvas (Visual Editor)
```

### Component Categories
1. **Core Components**: Generic functionality (loops, parsing, multi-provider integrations)
2. **Bundles**: Service-specific components grouped by provider (OpenAI, Anthropic, etc.)
3. **Legacy**: Deprecated components (hidden by default)

### MCP Integration Architecture
- **MCP Client**: Connect to external MCP servers for tool access
- **MCP Server**: Expose Langflow flows as tools for other applications
- **Connection Modes**: JSON config, STDIO (local), HTTP/SSE (remote)

## Installation

### Prerequisites
- **Python**: 3.10-3.13 (macOS/Linux) or 3.10-3.12 (Windows)
- **Package Manager**: `uv` (recommended)
- **Hardware Minimum**: Dual-core CPU, 2GB RAM
- **Hardware Recommended**: Multi-core CPU, 4GB+ RAM

### Quick Start (Python Package)
```bash
# Create virtual environment
uv venv langflow-env

# Activate environment
# macOS/Linux:
source langflow-env/bin/activate
# Windows:
langflow-env\Scripts\activate

# Install Langflow
uv pip install langflow

# Launch
uv run langflow run

# Access at http://127.0.0.1:7860
```

### Docker Installation
```bash
docker run -p 7860:7860 langflowai/langflow:latest
```

### Desktop Application
Download from https://www.langflow.org/desktop

**Note**: Desktop version lacks Shareable Playground and Voice Mode features.

## Use Cases

### 1. Building an AI Agent Flow
```
Chat Input → Agent (with Tools) → Chat Output
             ↓
         [Calculator, Web Search, URL Fetcher]
```

**Steps**:
1. Add Agent component and configure LLM provider (OpenAI, Anthropic, etc.)
2. Connect Chat Input/Output components
3. Attach tools by enabling "Tool Mode" on components
4. Add system instructions for specialized behavior
5. Test in Playground with real queries

### 2. Model Context Protocol Integration
**Scenario**: Agent with external MCP server tools

1. Navigate to **Settings → MCP Servers**
2. Add MCP server connection:
   - **STDIO Mode**: Local server (command + args)
   - **HTTP/SSE Mode**: Remote server (URL)
   - **JSON Config**: Direct configuration object

From enuno/claude-command-and-control, MIT.

More from enuno/claude-command-and-control