A GitHub project called OmniRoute offers a free MIT-licensed gateway that exposes 1,400-plus models from 375 providers through a single endpoint, according to its repository description.
A GitHub project called OmniRoute offers a free MIT-licensed gateway that exposes 1,400-plus models from 375 providers through a single endpoint, according to its repository description.
The repository cheaper-inference/OmniRoute on GitHub describes an AI gateway released under the MIT license at no cost. According to the project description, it routes requests to 375 providers through one endpoint and supports models including Kimi, Claude, GPT, Gemini, GLM, DeepSeek and MiniMax. The listing says the gateway works with coding tools including Claude Code, Codex, Cursor, OpenCode, Cline and Copilot, and ships as a desktop application and a PWA.
The excerpt includes several quantitative claims. It states the gateway covers 375 providers, of which 150 or more are free, and 1,400 or more models. It claims quota-aware auto-fallback that switches between providers, and a compression feature called RTK plus Caveman that saves 15 to 95 percent of tokens. It also lists support for MCP and A2A protocols and credits more than 900 contributors with building the project.
The claims come only from the repository description, with no benchmarks, documentation tests or third-party measurements cited in the source. The token-savings range of 15 to 95 percent is unusually wide and unexplained, and the free-tier counts may depend on provider terms that can change. Anyone evaluating it should check the current code, license terms and rate-limit behavior directly on GitHub.