Hello Umesh Narayanan,
Thank you for posting your question in the Microsoft Q&A forum.
Agentic AI in Azure AI Studio can indeed support multi-agent applications similar to LangGraph's multi-node workflows, but requires careful orchestration to reuse agents across invocations. Currently, the create_and_process_run method instantiates a new agent per call, which isn't ideal for stateful multi-agent systems.
To reuse agents across invocations:
- Leverage the thread_id to maintain conversation context between runs, ensuring continuity.
- Cache agent instances externally (e.g., Azure Cache for Redis) and reference them via a shared agent_id.
- Orchestrate via Azure Logic Apps/Functions to manage agent lifecycle and state transitions.
For true multi-agent coordination, wrap agents in a LangGraph-like workflow using:
- Durable Functions for stateful chaining.
- Custom middleware to route inputs/outputs between agents.
If the above answer helped, please do not forget to "Accept Answer" as this may help other community members to refer the info if facing a similar issue. Your contribution to the Microsoft Q&A community is highly appreciated.