You are deep in a Claude Code session and Claude Max runs out of quota. Normally you stop and wait. MaxRelay keeps you moving. It is a tiny local proxy that sits in front of Claude Code: while Claude Max is healthy everything runs on Claude exactly as usual, and the moment it is out of quota MaxRelay automatically switches your session to another AI model, then falls through an ordered chain of backups if one is busy, and translates the answer back so tool calls and streaming keep working. You never lose your place. It speaks the Anthropic API, so Claude Code works out of the box. Type /relay to switch models live, or /relay max to jump back to Claude. Zero dependencies, Node 18+, and your keys and traffic stay on your own machine. Powered by Nexus.
maxrelay.zip below and double-click to unzip. Open the maxrelay folder and double-click MaxRelay.command. It starts the proxy on http://localhost:4575 and opens the control dashboard. First launch: if macOS blocks it, right-click the file and choose Open. Needs Node 18+ installed.localhost:4575, paste an OpenRouter key (many models, including free ones) or a DeepSeek key (cheap and reliable, recommended for real coding). Without a key Claude Max still works normally, there is just nothing to fail over to. Keys stay on your machine../relay instead of claude. It auto-starts the proxy and behaves exactly like Claude Code. Prefer manual? export ANTHROPIC_BASE_URL=http://localhost:4575 then run claude./relay deepseek (or any model name) to force a model, and /relay max to return to Claude Max. The dashboard lists every available model, tap one to use it.