Bring your own key

You can bring your own Anthropic, OpenAI, Google or other provider key. Add it once, in the dashboard — not in the editor.

Adding a key

Go to Dashboard → AI → Credentials and add the key for your vendor. That is the whole process.

Models it unlocks appear in your editor's picker on the next refresh. No extension setting, no restart, no configuration file.

Why not in the editor

This is deliberate, and it is better than the alternative:

• One place, every app. The same key serves the extension, the CLI, the desktop app and the JetBrains plugin. Adding it four times is four chances to get it wrong.
• Keys never sit in editor state. They do not end up in a synced settings.json, in a workspace file you might commit, or in an extension's storage.
• Revoking works everywhere at once. Remove it in the dashboard and every app stops using it immediately.

The editor has no API-key field at all. There is nothing there to leak.

What changes when you use your own key

• You pay the vendor directly for the tokens, at their price.
• ToolsXpo charges a small share of what the call would have cost at the vendor, instead of token credits.
• You get access to models we do not fund — the catalogue is larger with a key than without one.

The model picker's default filter shows what you can call right now, which accounts for your credentials. A model that would need a key you do not have is not offered.

Adding your own models

You can also add models we do not carry: Dashboard → AI → Models. They are scoped to your workspace and appear in your picker like any other.

Rotation and failover

Keys can be rotated without downtime, and a key that starts failing is rested rather than retried on every request — so a thirty-second outage at a vendor does not become something you have to notice and undo.

See also

• Choosing a model
• AI credits