Connect Cherry Studio to Claude

Cherry Studio is a tidy desktop client for chatting with many models, and it speaks the native Anthropic format directly. Adding the AI Prime Tech gateway as an Anthropic-type provider gives you Claude in the same window as everything else you run, with one host and one key.

Open Model Services

Go to Settings and open Model Services. This is where Cherry Studio lists every provider it can talk to, and where you add new ones. Press the plus to add a provider and you are asked what type it is.

Choosing the type matters because it sets the wire format. For Claude on the gateway you want the Anthropic type, which targets the native /v1/messages surface rather than an OpenAI-style endpoint.

Because Cherry Studio keeps providers as a list, the gateway sits alongside any other services you have configured, and you switch between them per conversation. That makes it a comfortable home for a multi-model gateway: one provider entry can carry several Claude models that you pick from the menu as needed. Treating the gateway as one entry among several is what makes Cherry Studio a comfortable multi-model cockpit. You switch providers per conversation without leaving the app.

Pick the Anthropic type

Select Anthropic as the provider type and enable it. Cherry Studio then shows two fields that decide everything: the API Key and the API Host. Paste your gateway key into the key field.

Because the type is native Anthropic, Cherry Studio will send requests in the Messages format the gateway serves for Claude. There is no separate protocol toggle to worry about once the type is set.

If your gateway also exposes an OpenAI-compatible surface for other models, you would add those as a separate provider of the OpenAI type rather than mixing them into this one. Keeping the Anthropic-type entry dedicated to Claude keeps the wire format unambiguous and avoids cross-protocol confusion. Keeping protocols on separate provider entries is a small discipline that prevents a lot of confusion later. Each entry then has exactly one wire format to reason about.

Set the API Host correctly

The API Host is the field people most often get wrong. Enter the host root only, with no /v1 on the end, because Cherry Studio appends the version path itself. Adding /v1 yourself produces a doubled path and a failed request.

So the value is just the gateway host. If you have seen 404s or path errors after adding a provider, a stray /v1 in the API Host is almost always the cause.

A trailing slash on the host causes the same kind of breakage, so enter the bare host with nothing after it. When a request fails, the quickest diagnostic is to read back the API Host field and strip anything beyond the host name before looking anywhere else. When something fails, reading back this one field first will resolve most setup problems on the spot. A bare host with no suffix is what you are aiming for.

The trailing-hash lock

Cherry Studio supports a small but important trick: end the API Host with a # and the client uses that exact URL without appending anything. This is how you force a precise endpoint when auto-append gets in the way.

For a standard gateway you usually leave the host bare and let Cherry Studio append. Reach for the # only when your gateway needs an exact path the auto-append would otherwise mangle.

The # is worth remembering precisely because it is non-obvious — it is the escape hatch for any gateway whose path does not match Cherry Studio's default assumption. If you ever find the auto-appended /v1/messages does not line up with what your gateway expects, the # lets you spell out the full URL by hand. Filing the # trick away now means you will recognise the fix instantly if a path ever misaligns. It is the one override that handles any non-standard gateway path.

Choose a model and chat

Add the Claude models you want under the provider, using the ids the gateway serves such as claude-sonnet-4-5 for everyday chat and claude-opus-4-8 for harder reasoning. They then appear in Cherry Studio's model menu.

Start a conversation, pick a Claude model from the menu, and you are talking to the gateway. Because the provider is native Anthropic, features that depend on the Messages format behave as expected.

You can keep several conversations open against different Claude models at once, which is handy for comparing how Sonnet and Opus handle the same prompt. Because they all run through the one gateway provider, switching models is a menu choice rather than a reconfiguration. Side-by-side model comparison is one of the quiet pleasures of running everything through a single provider. The same prompt against Sonnet and Opus is one menu switch apart.

# Cherry Studio -> Settings -> Model Services -> + Add
# Provider type: Anthropic   (native /v1/messages)
# API Key:  <your gateway key>
# API Host: https://aiprimetech.io        (root only, NO /v1 -- it auto-appends)
#   end with #  ->  https://aiprimetech.io#   to lock the exact URL
# Model: claude-sonnet-4-5

Frequently asked questions

How do I add Claude to Cherry Studio?
Settings, Model Services, Add, choose the Anthropic type, paste your key, and set the API Host to {SITE} (root, no /v1).

Should the API Host include /v1?
No. Use the host root; Cherry Studio appends the version path itself. Adding /v1 yourself doubles the path and fails.

What does the trailing # do?
It locks the exact URL — Cherry Studio uses the host as-is and appends nothing. Use it only when you need a precise path.

Which provider type for Claude?
Anthropic. It targets the native /v1/messages surface the gateway serves for Claude.

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.