Claude Opus 5.5 in Cursor, Cline and Cherry Studio: Which API Type and Base URL to Use

Short answer

Choose the Anthropic type whenever the client offers it: prompt caching works, and on long conversations the input bill can differ by around ten times. Whether the base URL ends in /v1 depends on how each client builds its request path; what matters is that requests end up at https://api.teamorouter.vip/v1/messages (Anthropic) or https://api.teamorouter.vip/v1/chat/completions (OpenAI).

On TeamoRouter, Opus 5.5 costs roughly 30% of the list price (as of writing — the live numbers are on the pricing page) and has a 1M-token context. Create an API key in the console; the three clients below and Claude Code can all share one key.

Why the Anthropic type is worth choosing

TeamoRouter accepts both Anthropic format (/v1/messages) and OpenAI-compatible format (/v1/chat/completions), with the same models and prices. For Opus 5.5 the difference is caching:

• Anthropic format lets a request mark the system prompt, tool definitions and earlier messages for caching (cache_control). Whatever is sent again is billed at the cache-read rate — $0.20 per million tokens at list price for Opus 5.5, just 5% of normal input ($4).
• OpenAI format has no such field, so the context that a long conversation resends on every turn is generally billed as normal input.

An example: an agent session that carries about 100k tokens of context per turn, over 30 turns.

New content per turn and output cost the same either way. On TeamoRouter both cases are roughly 30% of list, so the ratio holds.

The client has to put the cache markers in the request. Claude Code does it automatically, and current Cherry Studio 2.x enables it by default for Anthropic-type providers.

Cherry Studio: use the Anthropic type

1. Click ⚙ in the top-right corner to open Settings, choose Model Services on the left, and click Add at the bottom of the provider list.
2. Name it anything, for example TeamoRouter Claude, and set the provider type to Anthropic. If you already have an OpenAI-type TeamoRouter provider for GPT models, create a separate Anthropic-type one for Claude.
3. Paste your TeamoRouter key as the API Key, and set the API URL to https://api.teamorouter.vip, without /v1. Cherry Studio adds the path itself; the preview under the field should read https://api.teamorouter.vip/v1/messages.
4. Models: click Fetch model list and select claude-opus-5-5, or add that ID manually.
5. Click the check button beside the API Key. Once the connection succeeds, turn on the switch in the top-right corner of the panel.

Step-by-step screenshots are in the Cherry Studio setup guide.

Cline: Anthropic, with a custom base URL

1. Click the Cline icon in the VS Code sidebar and open the settings in the top-right corner of the panel.
2. Set API Provider to Anthropic.
3. Paste your TeamoRouter key into the API key field.
4. Check Use custom base URL and enter https://api.teamorouter.vip/v1.
5. In the Model dropdown, pick Claude Opus 5.5. If it isn't listed, update Cline to the latest version first — Cline's Anthropic provider doesn't accept a typed model ID, only one from the list.

Here the URL does end in /v1, the opposite of Claude Code. That's because the current Cline Anthropic provider's default address is https://api.anthropic.com/v1, and it appends /messages directly. If you're on an older Cline and get a 404 after saving, remove /v1 and try again.

Thinking can't be turned off on Opus 5.5 (per Anthropic), so leave Cline's Adaptive Thinking off None; choose low if you want to save tokens.

The fallback is OpenAI Compatible: Base URL https://api.teamorouter.vip/v1, Model ID claude-opus-5-5. It works, but it's OpenAI format, so caching mostly doesn't apply and long tasks cost noticeably more — see the note in the Cline setup guide.

Cursor: OpenAI-compatible only

In Cursor's own-key settings, the custom base URL lives under OpenAI, so Opus 5.5 can only be connected in OpenAI-compatible format:

1. Press ⌘ + , (Ctrl + , on Windows and Linux) to open Cursor Settings and choose Models.
2. In the API Keys section, paste your TeamoRouter key into OpenAI API Key.
3. Expand Override OpenAI Base URL, enter https://api.teamorouter.vip/v1, and turn the switch on.
4. Add claude-opus-5-5 to the model list.
5. Select it from the model picker in Chat (⌘ + L) or Composer (⌘ + I).

Three limits to know first:

• Cursor's documentation states that your own API keys only work with chat models; Tab completion keeps using Cursor's built-in models.
• Per the Cursor setup guide, a custom key and base URL have no effect on the Free plan; you need a paid plan.
• It's OpenAI format, so caching mostly doesn't apply. For long agentic tasks in Cursor, it's cheaper to run Claude Code in Cursor's built-in terminal — setup is in Claude Opus 5.5 API: pricing and setup.

Common errors

"thinking.type.disabled" is not supported for this model or "thinking.type.enabled" is not supported. The client turned thinking off or set a fixed thinking budget. Anthropic states that thinking is always on for Opus 5.5 and is controlled only through effort. Set the client's thinking option back to default or adaptive, or pick an effort level.

tool_choice: type "tool" and "any" are not supported for this model. The client is forcing the model to call a tool. Per Anthropic, Opus 5.5 doesn't support forced tool use and only accepts auto and none. Turn off any "force tool use" option in the client, or update it to the latest version.

404. Usually the /v1 part: Claude Code and Cherry Studio leave it off, Cline and Cursor need it.

401. The key is wrong, or a space or line break was pasted along with it. Copy it again from the console.

Paying for usage

You top up a balance in the console and are charged only for tokens actually used — no subscription or monthly fee. When several clients and Claude Code share one key, they all draw on the same balance.

• Card — Visa, Mastercard and other major cards via Stripe.

FAQ

Can I use the Anthropic API Key field in Cursor? That field has no base URL override, so a key entered there is sent to Anthropic itself. To use TeamoRouter, use the OpenAI field.

Is the model ID claude-opus-5-5 or claude-opus-5.5? claude-opus-5-5, with hyphens.

Can GPT and Claude share one provider in Cherry Studio? Not recommended. Use the OpenAI type for GPT models and the Anthropic type for Claude models — two providers, which can use the same key.

Does the same task cost the same in every client? No. The per-token price is identical, but clients that use OpenAI format miss the cache, so the input side of long conversations costs much more, and the gap grows with conversation length.

Next steps

• Claude Opus 5.5 API: pricing and setup — model ID, prices, Claude Code configuration.
• Opus 5.5 vs Sonnet 5.5 vs Fable 5.1 — prices and how to split models in Claude Code.
• Opus 5.5 model page — specs and live pricing.
• Create an API key — one key for every client.