Robotics MCP Server

# TODO: Agent-to-Agent Handoff Integration (Frontend) This document outlines the frontend requirements for handling agent-to-agent behavioral delegation within the `robotics-webapp` chat interface. ## 🏁 Goal Provide a seamless user experience when the underlying AI persona shifts between a Generalist Assistant and a Robotics Specialist, including Human-In-The-Loop (HITL) approval. --- ## 💬 Chat Interface Enhancements - [ ] **Handoff Status Handler**: - Detect `HANDOFF_PENDING` status in LLM responses. - Display a "Handoff in Progress" animation/status indicator. - [ ] **Persona Visualization**: - Update the chat header/avatar dynamically when the active agent changes (e.g., from `Sandra-Generalist` to `Sandra-Robotics`). - Clear/Archive previous conversation bubbles if the context window resets. ## 👥 HITL (Human-In-The-Loop) Component - [ ] **Approval Dialog**: - Create a modal/inline component for `confirm_handoff()`. - Show the user: - Which agent is taking over. - What permissions/tools the new agent will have access to. - A summary of the context being transferred. - [ ] **Security Disclosure**: Show a warning if the target agent has "elevated" capabilities (e.g., physical robot control). ## 🎛️ State Management - [ ] **Context Refresh**: - Trigger a proactive `get_status` call to the `robotics-mcp` server after a handoff completes to ensure the UI state matches the new agent's understanding. - [ ] **Identity Storage**: - Ensure the OIDC/JWT token is passed in all headers during and after the handoff process. --- ## 📚 Reference See [AGENT_TO_AGENT_HANDOFF.md](file:///d:/Dev/repos/mcp-central-docs/docs/patterns/AGENT_TO_AGENT_HANDOFF.md) in the `mcp-central-docs` repository for the full pattern specification.