flo2 — LLM Gateway & Router | Cheapest, Fastest Models, One API Key - Showcase Project

flo2 is a large language model (LLM) gateway, router, and proxy for developers and engineering teams. It provides a single access layer in front of multiple model providers, so applications can call many models through one unified API instead of maintaining a separate integration for each provider. Rather than reselling tokens, flo2 connects to providers using keys supplied by the user and applies no markup to their rates; requests are sent directly to the providers. A single OpenAI- and Anthropic-compatible API key is used to reach models across providers including OpenAI, Anthropic, Groq, Cerebras, DeepInfra, and Gemini. Because the interface follows existing standards, adopting flo2 typically requires only changing the base URL in an existing integration. flo2 records usage data for each call, including token counts, throughput, and computed cost. This information is available for review, so spending and performance can be monitored per model and per provider. Response caching is available on an opt-in basis, with a configurable time-to-live. flo2 works as a drop-in replacement for the OpenAI Chat Completions, Responses, and legacy Completions endpoints and the Anthropic Messages API, including streaming responses. flo2 is intended for teams running LLM-powered products in production, particularly infrastructure-heavy, high-volume applications where model cost, reliability, and data handling matter. It is currently free during its Beta period.