What problem does it solve? Configuring MCP servers across Claude Code, Codex, Gemini, and OpenClaw is error-prone, and teams waste budget by routing every task to expensive models when free or local agents would succeed. ## Core Features & Use Cases - MCP Server Setup & Debugging: Register and verify MCP servers (gemini-mcp-tool, ai-cli-mcp) in Claude Code, Claude Desktop, and OpenClaw with JSON config templates and a verification checklist. - Cost-Aware Model Routing: Route coding tasks to ClinePass, lightweight triage to the OpenRouter free-model fallback chain, local work to ollama, and reserve Gemini strictly for large-context Gemini-Analyzer use-cases. - Parallel Worker Dispatch: Use ai-cli-mcp to run background CLI agents with PID tracking, absolute workFolder paths, and human approval gates before any writes. - Use Case: When an MCP server fails to appear in /mcp or a background agent hangs, follow the troubleshooting table to fix auth, JSON parse errors, or first-run prompt issues. ## Quick Start Ask the agent to set up and verify the ai-cli-mcp and gemini-cli MCP servers in Claude Code and route a coding task to the cheapest capable model.