Cursor
Cursor can send its OpenAI chat requests to an OpenAI-compatible address with your own key. With MartLLM as that address, Cursor's chat uses MartLLM models and your MartLLM balance.
Before you start
- A MartLLM account with credit, and an API key from
Dashboard → Keys (
ull_...).
Setup
Open Cursor Settings → Models.
Under OpenAI, paste your MartLLM key into the API Key field and save.
Turn on the OpenAI base URL override switch and enter:
https://api.martllm.com/v1Turn on Use API Key for OpenAI.
Add the MartLLM model id you want as a custom model (for example
deepseek/deepseek-v4.1-flash) and select it in the chat model picker.
What to expect
- Chat only. Cursor uses your own key for chat models. Tab completion keeps using Cursor's own models.
- The override applies to every OpenAI model in Cursor. While it is on, built-in
OpenAI model names (
gpt-*) are also sent to MartLLM, so pick a model id MartLLM serves. - Requests go through Cursor's servers. Cursor builds the prompt on its side and
then calls the base URL, so the address must be reachable from the internet.
api.martllm.comis.
Recommended models
Checked on 2026-10-30 through MartLLM.
| Model id | Context | Notes |
|---|---|---|
deepseek/deepseek-v4.1-flash |
1,048,576 | Fast; reasons before it answers |
z-ai/glm-5.2 |
1,048,576 | Strong at code |
google/gemini-2.5-flash |
1,048,576 | Reads images |
Check the key and URL from a terminal
If Cursor gives no clear error, test the same settings directly:
curl https://api.martllm.com/v1/chat/completions \
-H "Authorization: Bearer $UBERLLM_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"deepseek/deepseek-v4.1-flash","messages":[{"role":"user","content":"ping"}]}'
A 200 with choices[0].message means the key, URL and model id are right.
Common errors
| What you see | Cause and fix |
|---|---|
| Requests fail right after saving the key | Cursor says requests using a rejected key fail until you fix or remove it. Re-paste the key; check it with the curl above. |
404 with "page not found" |
The override URL must end in /v1 and nothing after it. |
404 model_not_found |
A built-in gpt-* name or a mistyped id was sent to MartLLM. Select your custom model. |
402 insufficient_credits |
Balance empty. Add credit under Dashboard → Credits. |