Spaces:
Running
Running
Commit History
-
feat: Simplify MCP responses and improve routing config UI Running
KISS PRINCIPLE - MCP Response Simplification: - Return just agent name ("gemini"/"aider"/"copilot"/"claude") - No more verbose JSON with redundant fields - Orchestrator only needs WHERE to route, not WHY or HOW - Reduces parsing overhead for external clients UI/UX Improvements: - Remove "Just like install.py!" text from config accordion - Add agent selection checkboxes (Gemini/Claude/Aider/Copilot) - Convert Radio buttons to Dropdowns for better UX - Auto-disable unselected agents in task category dropdowns - Auto-reassign task categories if selected agent disabled - Real-time updates when agent availability changes Technical Changes: - server.py: Simplified get_routing_guidance response - system_instructions.py: Updated CLAUDE.md format for new response - gradio_monitor.py: Enhanced task routing configuration UI Breaking Changes: - MCP tool response format changed from JSON to plain string - Old: {"decision": "DELEGATE_TO: gemini", ...} - New: "gemini" 🤖 Generated with [Claude Co…c77464a -
Initial public release: Multi-Agent MCP Delegation Server
A Model Context Protocol (MCP) server that enables intelligent task delegation across multiple AI coding agents (Claude, Gemini, Aider). Features: - Capability-based routing with automatic fallback - 10+ specialized task types (security, refactoring, code review, etc.) - Flexible configuration via YAML and Python install.py wizard - Hugging Face Space demo with interactive query tester - Comprehensive test suite and example workflows - Docker support for easy deployment Token efficiency: <1% context overhead (1,739 tokens) License: MIT8b02e7c