LiteLLM is an open-source AI gateway and LLM proxy that puts more than 140 providers and 1,800 models behind one OpenAI-compatible API key. It tracks who is driving spend, caps usage with budgets and rate limits before a request runs, and routes each call to the right model with fallbacks. The Open Source version is free to self-host forever, and an Enterprise tier adds SSO, SCIM, audit logs and support with custom pricing based on request capacity, never per token. A 30-day Enterprise trial key is available.
LiteLLM is built for platform teams that must open model access to a whole company without becoming a bottleneck. Developers get one API and day-zero access to new models, while security and finance staff get spend visibility by key, user, team and organization. Regulated companies appreciate that it self-hosts anywhere, including air-gapped environments, so data and keys stay on their own infrastructure. Individual developers and small teams can run the open-source proxy in minutes without a credit card. Testimonials come from engineers at NVIDIA, Netflix, Okta, Lemonade and AT&T.
Common uses include swapping the backend model through a configuration change instead of code changes, issuing scoped virtual keys to each team or project, and enforcing per-team budgets so one runaway job cannot drain a monthly allowance. Companies bring internal, fine-tuned and self-hosted models behind the same key as hosted ones, and reach agents and MCP servers through the same gateway. Monitoring teams scrape Prometheus metrics and keep request logs for audits. AT&T reported coding costs falling by as much as 56% after routing through LiteLLM. Enterprise buyers can purchase through AWS Marketplace.
