Unified API Gateway for AI Services
Access LLM, TTS, STT, and more AI services with a single API key.
Explore available APIs and upcoming services.
llm.apis.how
Multi-provider AI proxy. Supports text generation, chat, and agent mode.
chorus.how
Multi-LLM consensus system. Cross-validate with 4-6 AIs for higher reliability.
tts.apis.how
Convert text to natural speech. Supports multiple voices and languages.
stt.apis.how
Accurately convert speech to text. Supports real-time transcription and file upload.
apis.how is not a simple API proxy. Apps call an OpenAI-compatible API with virtual keys and model aliases, while provider accounts, keys, billing, quotas, incidents, fallback, and usage are absorbed by the operations layer.
Call contract
Apps use LiteLLM virtual keys and model aliases instead of raw provider keys or provider-specific model names.
Operating cost
We manage retries, failed requests, streaming disconnects, provider incidents, abuse protection, and support cost.
Account management
We operate provider consoles, billing accounts, API keys, regions, permissions, and key rotation as one layer.
Manage provider accounts and API keys for OpenAI, Google, Azure, Deepgram, Fish, Runway, and more by customer service.
Apps call aliases like `primary-chat` or `cheap-batch`; provider changes are handled in gateway settings.
Route requests to the right provider based on workload, cost, latency, and provider health.
Detect transient failures and switch to retry logic or alternate providers.
Normalize REST, gRPC, WebSocket, and streaming responses into service-friendly interfaces.
Convert provider-specific tokens, characters, seconds, and requests into unified usage records.
Manage limits and call rates by customer, tenant, app, and token.
Connect provider costs and customer usage to catch budget overruns and abnormal usage early.
Track success rate, latency, incidents, cost, usage, and audit logs from an operations dashboard.
An API management platform designed for developers
Access all services with one key. Fine-grained scope-based permissions.
Multi-layer rate limiting: global, per-service, and per-token.
Real-time call counts, token usage, and cost estimation monitoring.
Automatic provider switching on failure for high availability.
Restrict access by IP to prevent unauthorized usage.
Track all API call history and anomaly detection alerts.
Call AI APIs with just a few lines of code.