> ## Documentation Index > Fetch the complete documentation index at: https://docs.reasoningstack.ai/llms.txt > Use this file to discover all available pages before exploring further. # Agent Desk > Manage, monitor, and visualize multi-agent collaboration in real time. The Agent Desk lets you follow the full conversation and workflow between AI agents. The **Agent Desk** is your real-time mission control for all AI agents collaborating on your tasks in Reasoning Stack.\ Think of it as an interactive dashboard where you can track, inspect, and understand how sub-agents coordinate, share messages, and deliver results for your request. *** Agent Desk UI Light *** ## 🚦 What is the Agent Desk? * **Live agent cards**: See every AI agent working on your task, each with its own avatar, name, specialty, and real-time status. * **Workflow visualization**: Watch as agents receive tasks, process information, hand off messages, and return results. * **Message threads**: Track conversations between agents—see who said what, when, and to whom. * **Status indicators**: Know at a glance who’s busy, who’s done, and who needs more info. *** ## 🛠️ How it works ### 1. Agent Plan Creation When you submit a request, the system generates an **Agent Plan** using the `create_agent_plan` tool.\ This defines which agents (sub-agents) are needed, their roles, and the overall sequence. * Each agent gets assigned an avatar for easy visual tracking. * The **lead agent** acts as the coordinator. ### 2. Agent Communication * **Agents communicate by passing messages:**\ Each agent can send structured messages (with context, data, or instructions) to other agents as the workflow progresses. * **Message routing is dynamic:**\ Agents hand off tasks and context to each other based on the Agent Plan—there’s no fixed chain; agents can "talk" back and forth as needed until every sub-task is completed. * **Conversation is visible:**\ You can follow every exchange in the Agent Desk UI, so it’s clear how agents collaborate and which step each is handling. ### 3. Full Orchestration Loop * The process continues, with agents exchanging results and new instructions, until all sub-tasks are done and the lead agent delivers the final output back to you. ***