Modelmux is a cross-platform server designed to unify access and control over diverse AI models and external agents. It allows any MCP-compatible platform to dispatch, broadcast, and orchestrate tasks across providers such as Codex CLI, Gemini CLI, Claude Code CLI, Ollama, DashScope, and external A2A agents. With advanced features like smart routing (v4), exponential retry, cost tracking, and robust multi-agent collaboration patterns (review, consensus, debate), Modelmux ensures optimal utilization of each model's unique strengths, automatic failover, and efficient resource management for complex AI-driven workflows.
主な機能
01Smart Routing v4 with keyword, history, benchmark, and user feedback scoring
02Multi-Model Orchestration for dispatch, broadcast, and collaborative tasks
03Cost Tracking with token usage extraction and per-model estimation
04A2A Protocol support for agent-to-agent interoperation via HTTP server/client
05GitHub Actions for automated PR code review and integration
060 GitHub stars
ユースケース
01Automating code review and debugging processes across multiple AI models via GitHub Actions.
02Facilitating iterative multi-agent collaboration patterns (e.g., implement, review, revise) for complex software development tasks.
03Optimizing AI task execution by intelligently routing requests to the best-suited model based on content, cost, and availability.