What is OpenFlow?

⚠️ NOTICE: This protocol is currently under active development and is not fully complete. Specifications and features may change without notice.

Abstract

The OpenFlow Protocol is a comprehensive, standardized specification for defining and executing AI-powered workflows through declarative JSON structures. This protocol enables the orchestration of complex AI operations involving Large Language Models (LLMs), vector databases, document processing, and control flow mechanisms in a vendor-agnostic, portable, and auditable manner.

The protocol serves as the foundation for building interoperable AI systems that can be shared, validated, and executed across different environments and implementations while maintaining consistency and reliability.

OpenFlow enables you to build complex AI workflows by chaining together different types of nodes:

LLM Nodes: Generate text, analyze images, process documents
Vector Database Nodes: Store and search embeddings
Document Processing: Split PDFs into pages, extract text and images
Control Flow: Conditional logic, loops, variable manipulation
Embedding Nodes: Generate text embeddings
MCP Integration: Model Context Protocol support for external tools and data sources

What is OpenFlow?

Abstract

Example Orchestration

Basic LLM

Model Context Protocol Integration

Vector Database Insert with Text Embedding

Vector Search

On this page