GPT-Load GPT-Load

Consolidate API keys, subscription accounts, traffic scheduling, fault handling, request logs, and usage statistics into a single entry point. Why choose GPT-Load Applications only need to configure one address and one AccessKey. All backend providers, accounts, credentials, models, and routing strategies are managed entirely within the administrative interface. - Unified Entrance, Native Protocol Retention — Centrally manage official APIs, cloud platforms, model services, and compatible relays; clients continue using native OpenAI, Anthropic, or Gemini interfaces without any code modifications. - Unified Management of API Keys & Subscriptions — Subscription channels like Codex, Claude, Antigravity, and Grok share the same credential management, scheduling, and health systems as API Key channels. - Built-in Scheduling & Fault Isolation — Multi-credential scheduling, automatic weighting, retries, cooldowns, blacklisting, and session affinity reduce the impact of individual credential overload or failure. - Observable, Easy to Deploy, Data Self-hosting — Provides health monitoring, routing, logs, usage, and cost estimation; a single Go binary with an embedded management UI, supporting SQLite, MySQL, PostgreSQL, and local credential encryption. Support Scope Client Protocols | Protocol | Main Endpoint | | ----------------------- | -------------------------- | | OpenAI Chat Completions | POST /v1/chat/completions | | OpenAI Responses | /v1/responses & resource paths | | OpenAI Images | POST /v1/images/... | | OpenAI Embeddings | POST /v1/embeddings | | Anthropic Messages | POST /v1/messages | | Gemini | /v1beta/models/... | Each channel explicitly declares its executable protocols and capabilities. GPT-Load performs conversions between supported capabilities but is not a universal converter for arbitrary protocols or JSON. In the initial phase, Embeddings are only provided via native OpenAI-compatible Wire through OpenAI, OpenRouter, and OpenAI Compatible API Key channels; they are not supported for subscription channels or protocol cross-conversion. AccessKeys without protocol filters will allow all enabled protocols by default and will gain Embeddings access after upgrading; for least-privilege deployment, please explicitly configure protocol filters. Built-in Channels - Official & Cloud Platforms: OpenAI, Anthropic, Gemini, xAI, Azure OpenAI, AWS Bedrock, Google Vertex AI - Popular Model Services: DeepSeek, Moonshot AI, SiliconFlow, Zhipu AI, Alibaba, Volcengine, OpenRouter, Groq - Subscription Channels: Codex, Claude, Antigravity, Grok - Custom: OpenAI Compatible (any compatible relay)