Skip to content

Migration Guides

Migrate from other AI frameworks to LionAGI.

Available Migration Guides

Choose Your Migration Path

Coming from LangChain/LangGraph?LangChain Migration Guide - Reduce complexity by 90%
Coming from CrewAI?CrewAI Migration Guide - Keep your crew concepts, gain performance
Coming from AutoGen?AutoGen Migration Guide - Replace conversations with graphs

From LangChain

Escape LCEL complexity and gain true parallel execution. Perfect if you're tired of complex state management and want cleaner abstractions.

From CrewAI

Keep your agent crew concepts while gaining LionAGI's superior orchestration. Ideal if you like CrewAI's simplicity but need better performance.

From AutoGen

Replace unpredictable agent conversations with deterministic workflow graphs. Great if you want AutoGen's multi-agent power with production reliability.

Migration Philosophy

Zero-Disruption Migration: Keep your existing code and gradually adopt LionAGI orchestration.

# Your existing workflow runs unchanged
async def existing_workflow(input_data):
    return await your_current_implementation(input_data)

# Orchestrate with LionAGI
builder.add_operation(operation=existing_workflow)

Migration Strategies

Gradual Adoption

  1. Wrap existing workflows as custom operations
  2. Add LionAGI orchestration around them
  3. Gradually convert individual components
  4. Gain orchestration benefits without disruption

Hybrid Workflows

  • Mix existing framework code with native LionAGI operations
  • Coordinate multiple frameworks in single workflow
  • Best of all worlds approach

Full Migration

  • Translate framework patterns to LionAGI equivalents
  • Leverage LionAGI's superior orchestration capabilities
  • Gain performance and simplicity benefits

Why Migrate?

  • Parallel Execution: LionAGI runs operations concurrently by default
  • Simpler Code: Less boilerplate, cleaner abstractions
  • Framework Agnostic: Orchestrate any tool, not just one ecosystem
  • Production Ready: Built-in monitoring, error handling, performance control