gpt-load

GPT-Load is a Go-based reverse proxy server that sits in front of multiple AI provider APIs, including Anthropic Claude, OpenAI, and Google Gemini, routing requests while managing pools of API keys across each provider. It connects to Claude through the native Anthropic API format, forwarding requests transparently to upstream endpoints without altering the request or response structure. Key features include weighted load balancing across multiple upstream addresses, automatic key blacklisting when failures are detected with scheduled recovery, and hot-reload of configuration without service restarts. A Vue 3 web dashboard at the management interface handles group configuration, real-time request logging, and health monitoring. Deployment options range from a single Docker container backed by SQLite to a distributed leader-follower cluster requiring MySQL or PostgreSQL and Redis for coordination. The tool targets developers and enterprises operating at scale who need to aggregate many API keys across providers, reduce rate-limit exposure, and maintain availability through automatic failover rather than manual key rotation.

What people ask about gpt-load