Cline with teamToken

Cline has no config file: the provider is chosen in the extension settings and filled in with three fields. Setup takes a minute, and the one place that trips people up is the model name, typed as free text and checked by nothing.

What the tool is

Cline is a VS Code extension that edits code through an agent and reads the project as it goes. Any provider works as long as it speaks the OpenAI-compatible protocol.

A note for anyone arriving from Roo Code: the Roo team officially points former users at Cline, so the setup below applies to them too.

Three fields, no file

Settings → API Provider: "OpenAI Compatible"

  Base URL:  https://api.teamtoken.store/v1
  API Key:   <your teamToken key>
  Model ID:  gpt-5.6-sol

The provider type has to be OpenAI Compatible rather than "OpenAI". The first asks for a base URL; the second goes to OpenAI's address, where our key is not valid.

Model ID is typed by hand, and that is the risk

The field takes any string. No list is fetched, no autocomplete appears, and a wrong name surfaces as a provider refusal rather than as a typo. Take the name from the catalog rather than from memory:

curl https://api.teamtoken.store/v1/models \
  -H "Authorization: Bearer $TEAMTOKEN_KEY"

Names come back without a provider prefix. That is exactly the form to paste into Model ID.

Check before opening the extension

When something does not work, the useful thing to know is whose side to look at. One request to our endpoint answers that in a second.

curl https://api.teamtoken.store/v1/chat/completions \
  -H "Authorization: Bearer $TEAMTOKEN_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.6-sol","messages":[{"role":"user","content":"ping"}]}'

An answer means the key is alive, the balance is not empty and the gateway is reachable, so whatever is wrong lives in the extension settings rather than in the connection, and the provider type and Model ID are where to look. A refusal means the opposite, and opening VS Code will not help.

Which models to use

Spend caps per key

Cline reads project files into context, so it spends more than a chat-shaped tool and the spend grows with the repository. A separate key with a ceiling makes sense here.

When it does not work