Import agents from any framework. Enhance with enterprise features. Export to any target. No vendor lock-in, ever.
Built agents in CrewAI, LangGraph, or AutoGen? Import them directly.
Add SSO, audit logging, guardrails, and monitoring automatically.
Need to run on-premise with a specific framework? Export production-ready code.
Your agents are yours. Take them anywhere, anytime. We're the platform, not the cage.
Full bidirectional import/export — from Python source or from declarative YAML/JSON config.
Handoffs & guardrails
Role-based crews
Stateful workflows
Absorbed into MAF · import to migrate off
Tool chains
Claude API native
Absorbed into MAF · import to migrate off
NLP pipelines
RAG & indexing
Type-safe agents
Prompt optimization
Successor to AutoGen + SK
GCP-native, MCP + A2A
Model-driven, Bedrock-native
TypeScript-first agents
Declarative interchange format
Request a framework
Import/export keeps your agent definitions yours. These standards keep the runtime around them open too — every one of them is implemented in the platform, not on a roadmap.
Model Context Protocol — tools and data, spec 2026-07-28
Agent-to-agent, with a published agent card
SKILL.md, with progressive disclosure
The instruction convention, read and written
Open Agentic Schema Framework 1.1.0
OpenTelemetry semantic conventions for traces
Agents are also defined in configuration, not only in code. MeetLoyd reads
declarative YAML and JSON definitions directly
— a repository whose agents live in agents.yaml imports
without a line of Python to parse.
MeetLoyd concepts translate to each framework's native patterns.
| Framework | Agents | Teams | Tasks | Workflows | Tools | Handoffs |
|---|---|---|---|---|---|---|
| OpenAI Agents SDK | ✓ | — | — | ✓ | ✓ | ✓ |
| CrewAI | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| LangGraph | ✓ | — | — | ✓ | ✓ | ✓ |
| AutoGen | ✓ | ✓ | — | — | ✓ | — |
| LangChain | ✓ | — | — | ✓ | ✓ | — |
| Anthropic Claude | ✓ | ✓ | — | ✓ | ✓ | ✓ |
| Semantic Kernel | ✓ | — | — | ✓ | ✓ | — |
| Haystack | ✓ | — | — | ✓ | ✓ | — |
| LlamaIndex | ✓ | — | — | ✓ | ✓ | — |
| Pydantic AI | ✓ | — | — | — | ✓ | — |
| DSPy | ✓ | ✓ | — | ✓ | ✓ | — |
Each framework excels at specific patterns. We support them all.
Complex multi-step processes with conditional routing and cycles.
LangGraphAgents with distinct roles collaborating on tasks.
CrewAIAgents discussing and debating to reach conclusions.
AutoGenAgents in Microsoft ecosystems with Azure services.
Semantic KernelAgents with validated, typed responses.
Pydantic AIRetrieval-augmented generation with document indexing.
LlamaIndexProgrammatic prompt tuning and optimization.
DSPyDirect Anthropic API with native tool use.
Anthropic ClaudeImport, enhance, export. It's that simple.
# Import from AutoGen
from meetloyd import import_agents, export_agents
autogen_config = import_agents(
code=open("autogen_agents.py").read(),
framework="autogen"
)
# Enhance with MeetLoyd enterprise features
for agent in autogen_config.agents:
agent.guardrails = [pii_filter, content_safety]
agent.memory = {"type": "vector", "enabled": True}
agent.audit_logging = True
# Export to LangGraph for production
langgraph_export = export_agents(
agents=autogen_config.agents,
teams=autogen_config.teams,
framework="langgraph",
format="code"
)
# Generated files ready to deploy
for file in langgraph_export.files:
print(f"Generated: {file.path}")Import, enhance, and export across all 16 frameworks.