About LiteLLM
LiteLLM is an open source AI gateway that puts many model providers behind one OpenAI-style API. It lets platform teams control spending, routing and access from a single place.
What it does
LiteLLM accepts requests in one format and sends them to the model provider you choose. It can track and cap spending, so a team or key cannot exceed its budget. Routing rules decide which model handles each request. The gateway also supports MCP tools and agents, and it can be self-hosted on your own infrastructure, including air-gapped networks. A quick start script is provided for macOS, Linux and Windows, and a Python SDK is available.
Key features
- Route requests to many model providers through one API
- Track and cap spending by team or key
- Route each request to the model you choose
- Self-host the gateway on your own infrastructure
- MCP support for tool access
Who it's for
The gateway suits platform, backend and infrastructure teams that run several model providers and need one control point. A single developer calling one model may not need a gateway.
Pricing and good to know
The open source project is free to self-host. An enterprise tier with extra support and governance features is offered separately. Self-hosting means your team runs the service, keeps it updated and handles uptime, while model provider bills still apply.
Highlights
- Route requests to many model providers through one API
- Track and cap spending by team or key
- Route each request to the model you choose
- Self-host the gateway on your own infrastructure
- MCP support for tool access
