Hi everyone,
I’m building a multi-agent copilot in MindStudio with a Main.flow that contains a Chat block as the conversational hub. When the user expresses a specific intent, a Dynamic Tool in the Chat block jumps them to a specialized sub-flow (e.g., prospector.flow, copywriter.flow).
Each sub-flow also starts with a Chat block.
The problem: When the user lands in the sub-flow’s Chat block, the conversation history from Main.flow is gone. The sub-flow Chat block has no awareness of what was said before the jump. The user experience feels like starting a completely new conversation.
I am passing {{user_intent}} (the last message variable from Main.flow’s Chat block) into the sub-flow’s System Prompt, which helps with context, but the full conversation history is not available.
Is there a recommended pattern for building multi-agent architectures where the user feels like they are in one continuous conversation, even when jumping between sub-flows?
Is the correct approach to keep all conversation inside a single Chat block in Main.flow and use sub-flows only as linear tool workflows (Generate Text + User Input), rather than having Chat blocks in sub-flows?
Or should each specialized agent be a separate published app rather than a sub-flow, with the main app acting as an orchestrator that links to them?
Thanks,
SP