← All guides

How to use Claude Sonnet 5.5 in Claude Code

Claude Sonnet 5.5 is Anthropic's own model, so this is the easy case: Claude Code runs on it natively, no proxy and no endpoint tricks. Here's how to point Claude Code at it, what to check, and a cleaner path if you also want other models a click away.

TL;DR: Claude Sonnet 5.5 is an Anthropic model, so Claude Code speaks to it natively. You don't need a translating proxy or a custom ANTHROPIC_BASE_URL. You just need a valid Anthropic key (or your Claude plan) and the right model name selected. If you also want to flip to GPT, DeepSeek, Kimi or a local model without reconfiguring, use an agent that treats the model as a setting.

The easy case: Sonnet 5.5 is already native

Unlike routing Claude Code to a third-party model, there's nothing to translate here. Claude Sonnet 5.5 is Anthropic's own model, served over the Anthropic Messages API that Claude Code was built for. That means no proxy, no OpenAI-compatible shim, and usually no custom base URL, you authenticate with Anthropic and pick the model.

The config

What to set

  1. AuthSign in with your Claude plan, or set ANTHROPIC_AUTH_TOKEN (ANTHROPIC_API_KEY) to a valid Anthropic key. No custom ANTHROPIC_BASE_URL needed for Anthropic's own models.
  2. ModelSelect the Sonnet 5.5 model, using the exact model string Anthropic's API exposes.
  3. Nothing elseBecause it's native, there's no proxy to run and no endpoint translation to maintain.
Model names, versions, exact pricing and install commands change frequently, always check the provider's official documentation for the current details.

Why people pick Sonnet 5.5 for coding

Native fit

It's Anthropic's own model, so Claude Code's tool-calling and agent loop are tuned for exactly this family.

Balance

Sonnet lines aim for a strong quality-per-dollar point between the cheap and the flagship tiers.

No moving parts

No proxy to keep updated and no endpoint drift, the simplest setup of any model.

The cleaner option: Sonnet 5.5 plus everything else

The native setup is as easy as it gets, but Claude Code is single-provider by design. The moment you want to drop to a cheaper model for boilerplate, or fall back when your Claude plan is rate-limited, you're back to editing env vars and restarting. An agent where the provider, base URL, key and model are all settings lets you keep Sonnet 5.5 as your default and switch out on demand.

What that looks like with Wuwei

Sonnet 5.5 in the model list

Pick it in the UI; Claude's other tiers, GPT, DeepSeek, Kimi, Qwen and OpenAI-compatible endpoints sit alongside.

One-click switch

Move between Sonnet 5.5 and a cheaper or local model mid-task without editing config or restarting.

Free, open source, local

MIT-licensed, runs on your machine, reads your real files; you only pay your own provider usage.

FAQ

Does Claude Sonnet 5.5 work with Claude Code natively?

Yes. Sonnet 5.5 is an Anthropic model served over the Anthropic Messages API that Claude Code is built for, so it works natively, no translating proxy and no custom endpoint. Authenticate with your Claude plan or Anthropic key and select the model.

Do I need a proxy to use Claude Sonnet 5.5 in Claude Code?

No. A proxy is only needed for non-Anthropic models (like OpenAI's GPT or xAI's Grok). Sonnet 5.5 is Anthropic's own, so Claude Code talks to it directly.

What model name do I select for Sonnet 5.5?

Use the exact model string Anthropic's API exposes for Sonnet 5.5. The precise id is being verified, check Anthropic's model docs for the current value before scripting it.

How do I switch between Sonnet 5.5 and a cheaper model?

Use an agent where the model is a setting. Wuwei is free and open source, keep Sonnet 5.5 as your default and switch to DeepSeek, Kimi or a local model in one click instead of editing env vars and restarting.

Running Claude Code on Sonnet 5.5 is the no-friction case, it's native. If you also want cheaper or local models a click away for when you hit limits, Wuwei is a free, open-source agent that puts Sonnet 5.5 and everything else in one model list, running locally on your real files.

Windows · macOS · Linux · free, no login