Kilo Code with teamToken

Kilo Code supports custom providers, and teamToken goes in as a plain OpenAI-compatible one. Six fields in settings, one of which you will not have to fill: the tool fetches the model list itself.

What the tool is

Kilo Code is a VS Code extension that writes and edits code through an agent. You can add your own provider rather than waiting for a service to appear in its built-in list.

Six fields in settings

Settings, the providers section, add a custom one. The values are these.

The provider type has to be OpenAI Compatible. That is not cosmetic: it decides which protocol the tool speaks, and another type sends requests in the wrong shape.

The model list fills itself

Given a base URL and a key, Kilo calls GET /v1/models and populates the list from the reply. That route is live here and returns the whole text catalog, so no model names get typed by hand and none get mistyped either.

You can confirm the list will arrive without leaving the terminal:

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

If a list comes back here, it appears in the interface too. An empty reply or a 401 means the key is the problem rather than the extension's settings.

Which models to use

gpt-5.3-codex for editing code, gpt-5.6-sol for working out the task and planning, gemini-3.1-pro for a second opinion. They switch inside the interface, from that same fetched list, so trying a different one on the same task costs a dropdown rather than a round trip through settings.

The same key also reaches image and video models, which is worth knowing if the project you are editing needs assets as well as code. Those live on separate endpoints rather than inside the editor, but nothing else has to be set up for them.

Spend caps per key

An agent inside an editor slides into long request runs easily, especially on a large repository. Each key carries its own ceiling, so the one you hand the extension cannot take the whole balance with it.

When it does not work