← Help Center· Cover-up plan

General Compatibility Guidelines

Any tool that supports Custom OpenAI Compatible Interfaces can reach us. The tool you use is not in the list above? Just follow these general instructions.

1. Determine if your tool can be picked up first

Look for these keywords in its settings, and if any one of them appears, it means that it can be picked up:

Keyword	Description
Custom Interface/Custom Provider	You can create a new one of your own
OpenAI Compatible/OpenAI Compatible	Protocol Matching
Base URL/API Address/Interface Address	Changing the address is enough
Reverse proxy/proxy	Some tools use this as an entrance.

If you can only select the built-in "OpenAI/Anthropic/Gemini" without a custom entrance—— Then this tool cannot pick us up. Tell us the tool name and we'll see if there are alternatives.

2. Four values, alignment seating

No matter what the interface is called, the essence is these four items:

General filling method

API Addresshttps://sccclm. com/v1

KeyYour sk- * key

Model Namenewcclm-fast

ProtocolOpenAI compatible (do not select Anthropic)

3. To which floor should the address be filled?

This is where things go wrong the most. Different tools have different expectations for "interface address". Depends onWill it make up the path itself?：

Tool Behavior	You fill in	Examples
It makes up for itself`/v1/chat/completions`	Fill in the domain name only	`https://sccclm.com`
It only replenishes`/chat/completions`	Fill in to`/v1`	`https://sccclm.com/v1`
It requires a full endpoint	Complete	`https://sccclm.com/v1/chat/completions`

How do I know which one? Try it once and you'll know. Fill inhttps://sccclm.com/v1; if 404 is reported, replace it withhttps://sccclm.com; If 404 is reported, fill in the complete endpoint. You can definitely try it out within three times, no need to check the documentation.

4. Don't match it like this

Do not overlap/v1

# Error - will become /v1/v1/...
https://sccclm.com/v1/v1

Do not enter the real model name

# Error - we only recognize our stall name
gpt-4o
claude-3-5-sonnet
deepseek-chat

# Correct
newcclm
newcclm-fast
newcclm-coding
newcclm-free

Do not select the Anthropic protocol

Some tools will let you choose "OpenAI format" or "Anthropic format". Always choose OpenAI - if you choose another one you won't be able to connect.

Don't write keys into prompts or conversations

The key is only entered in the settings. Writing into conversations is both useless and revealing.

5. Test it

After matching, first use this to confirm that the service itself is working:

curl https://sccclm.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
-d '{"model":"newcclm-fast","messages":[{"role":"user","content":"Hello"}]}'

Results	Description
Returns JSON with content	No problem on the server side. → Next, check the tool settings.
`Invalid token`	The key is incorrect, it has nothing to do with us
`model not found`	Model name misspelled
`404`	The address hierarchy is wrong, see Section 3
Unable to connect/Timeout	Network problems

6. Give us the name of the tool if needed

Tell us which tool you are using and we will add a dedicated configuration instructions. It is enough to provide the tool name and a screenshot of the settings page (encrypted key).

The three most common scenariosIn fact, it is easy to solve: The address level is filled in incorrectly (see Section 3), the model name is filled in with the real model name, and the key is not copied completely. If you check yourself according to these three items first, most of the problems will disappear.

Back to Help Center · FAQs · Cherry Studio