Otari is Mozilla AI’s open-source, OpenAI-compatible LLM gateway that you run yourself. One endpoint fronts more than 40 model providers, with virtual keys, budgets and usage tracking built in, positioning it as a self-hosted alternative to hosted routing services. It has about 714 stars on GitHub and is under very active development in 2026.
Otari sits between your applications and model providers: it authenticates requests, keeps provider credentials behind the gateway, enforces budgets before dispatch, and records usage afterwards. Applications can use its OpenAI-compatible and Anthropic-compatible endpoints, so existing SDKs work by changing a base URL. A web dashboard manages providers, models, routing policies, keys, members and budgets, and routing supports failover, weighting and learned selection between providers. Optional components add code execution, web search, MCP tools and inference guardrails. Deployment is via Docker Compose with PostgreSQL, or a one-click Railway template.
Under the hood, provider calls are normalized through any-llm, Mozilla AI’s companion library, and Otari is developed as part of that broader open-source AI stack rather than as a standalone experiment.
Otari is free and open source under the Apache-2.0 license. You run it on your own infrastructure and pay model providers directly for usage; there is no per-seat fee or hosted subscription required.
Otari is for platform and product teams that call multiple LLM providers and need centralized keys, budgets and visibility without handing that control to a third-party router. It suits organizations with compliance requirements that demand a self-hosted gateway.