Putting Unlimited Claude Behind LiteLLM

LiteLLM concentrates the traffic of many apps into one endpoint, so it carries the densest, least predictable request stream in your stack. Putting a flat-rate unlimited Claude plan upstream is the cleanest way to make that aggregate cost a fixed number.

Configure model_list

Each config.yaml entry pairs a client-facing model_name with litellm_params. The model field's prefix sets the wire protocol — openai/<model> for the gateway's OpenAI-compatible surface (api_base with /v1), or anthropic/<model> for native Messages (root api_base).

Point api_base at the unlimited gateway and read the key via os.environ/<VAR> so it loads at proxy start.

The suffix gotcha

anthropic/<model> auto-appends /v1/messages to the root; disable that with LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX=true for custom paths. openai/<model> appends /chat/completions to a /v1 base. Keep the prefix aligned with the gateway surface or you get 400s.

The client-facing model_name stays arbitrary, so your apps can keep their own naming while LiteLLM maps to the gateway.

LiteLLM in front of Claude Code

A favourite setup is LiteLLM as a local proxy that Claude Code points at — 'litellm claude code proxy'. The proxy exposes one stable endpoint, you swap upstreams in config.yaml, and the unlimited plan covers all of it.

Searches for 'litellm claude max' reflect users wanting heavy Claude usage without the meter. A flat-rate gateway upstream is exactly that.

Why concentrated volume wants flat-rate

Per-token billing on the proxy's combined stream is the hardest cost in the stack to forecast. Flat-rate unlimited turns it into a single line item.

You can fan many teams and apps through one LiteLLM endpoint and keep the bill fixed for the period.

# config.yaml
model_list:
  - model_name: claude-sonnet-4-5
    litellm_params:
      model: openai/claude-sonnet-4-5
      api_base: https://unlimited.aiprimetech.io/v1
      api_key: os.environ/AIPRIME_KEY
# litellm --config config.yaml

Frequently asked questions

How do I route LiteLLM to unlimited Claude?
Add a model_list entry: model: openai/claude-sonnet-4-5, api_base {SITE}/v1, api_key via os.environ. Or anthropic/<model> with the root api_base.

openai/ or anthropic/ prefix?
openai/ for the OpenAI-compatible surface (base with /v1); anthropic/ for native Messages (root + /v1/messages).

Stop LiteLLM appending /v1/messages?
Set LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX=true and give the full path.

Why unlimited behind LiteLLM?
The proxy concentrates many apps' traffic into the densest stream in your stack — flat-rate unlimited makes that aggregate a fixed cost.

AI Prime Tech is an independent API gateway. It is not affiliated with, endorsed by, or a reseller of Anthropic. Claude and related model names are trademarks of their respective owners.