Any way to define agent via LangGraph or Strands?

[austinmw]

Hi, I assume this is not possible, but is there any extensibility or future consideration for allowing users to define the agent harness using popular orchestration frameworks like LangChain/LangGraph or Strands?

There's a lot of things that these frameworks simplify or have ecosystems built around, e.g.

• Predefined tool libraries
• Middleware (hooks)
• Conversation management
• Simplified deployment (Bedrock AgentCore)
• Tool and agent communication protocol implementations (MCP, A2A, etc.)
  etc.

I think needing to create an agent using a custom harness outside of these common frameworks could be a big barrier to developers training agents. Here's an example of what I'd love to be able to do with this library: https://github.com/OpenPipe/art-notebooks/blob/main/examples/langgraph/art-e-langgraph.ipynb