Cherry Studio lets you use several model providers in one desktop client. A custom AI API setup needs only three values: endpoint, API key, and model ID. Most failures come from entering too much of the endpoint path, pasting an incomplete key, or retaining an old model entry.
This guide uses Zivv's OpenAI-compatible endpoint.
Prepare the Account
Before opening provider settings:
- Sign in to Zivv
- Create a key on the API Keys page
- Choose a current model in Model Hub
Name the key after the client and device, such as “Cherry Studio - Laptop.” This makes usage easy to identify and lets you revoke one device without affecting other tools.
Add an OpenAI-Compatible Provider
Create a custom OpenAI-compatible provider and enter:
| Field | Value |
|---|---|
| API endpoint | https://zivv.pro/v1 |
| API key | The sk- key from the console |
| Model | An exact ID copied from Model Hub |
Stop the endpoint at /v1. Do not enter the full chat completions path because Cherry Studio appends the operation path itself.
Add and Select Models
Refresh the model list if the client supports discovery. When adding one manually, the value must exactly match the API model id.
Choose by workload:
- Coding, code review, and long documents: a current Claude model
- General work and OpenAI-compatible workflows: a current GPT model
- Large source material, images, or multimodal tasks: a current Gemini model
Model catalogs change. Current names and pricing belong in Model Hub, not in a permanent recommendation inside a setup guide.
Run a Connection Test First
Use Cherry Studio's connection test before enabling advanced features. If no test is available, start a new conversation and ask the model to “Reply OK.”
Only after that succeeds should you test web access, tools, images, and long context. This separates a base configuration failure from a model capability mismatch.
Fix a 401 Response
A 401 means authentication was not accepted:
- Copy the complete key again and remove spaces
- Confirm it is an API key, not the account password
- Remove old keys from duplicate provider profiles
- Fully restart Cherry Studio
Use a direct model-list request to verify the key separately. The 401 troubleshooting guide gives the complete sequence.
Fix a 404 or Missing Model
Check these first:
- The endpoint is
https://zivv.pro/v1 - The model ID comes from the current model list
If the issue remains, confirm that the key group permits the model and remove cached model entries. See the 404 model guide for a minimal test request.
Use Separate Keys Across Devices
One key per device is easier to control than one shared key:
- Usage identifies the device
- A lost device requires revoking only one key
- Each device can have a separate limit
- Server, Claude Code, and Codex keys stay unaffected
Teams should also issue one key per member and assign budgets through Teams. Never place a primary account key in a shared document or chat channel.
Protect the Key
Cherry Studio is a local client, but its key still acts like a password. Do not screenshot it, publish exported configuration, or commit it to Git. If exposure is possible, revoke it and create a replacement immediately.
Final Check
A complete setup has:
- A dedicated, clearly named key
- Endpoint
https://zivv.pro/v1 - An exact current model ID
- One successful minimal conversation
- A corresponding record in Usage
First-time API users can begin with the AI API quickstart. Protocol, SDK, and client details are maintained in the Zivv documentation.