How Do I Run OpenCode Through the Gateway?

OpenCode includes Vercel AI Gateway in its provider list. Once the connection is saved, model selection happens in the same terminal interface you already use.

Quick Answer

Connect OpenCode to Vercel AI Gateway once, then open /models and choose a model from the Gateway catalog. Enter the credential during setup, keep it outside the repository, and use a dedicated key with a budget.

Outcome

Select a Gateway model in OpenCode and complete one read-only task without exposing the credential.

Fast Track

Connect OpenCode to Vercel AI Gateway off camera or outside shared terminal history
Open /models and select a Gateway model
Run a small task in a throwaway repository

Hands-on exercise

Start with a throwaway repository containing demo.ts. Complete the /connect flow privately because it opens credential entry.

After the connection is saved, open the model picker:

/models

Choose Vercel AI Gateway, then select an available model. The exact catalog changes, so use the picker rather than copying an old model ID from a screenshot.

Try It

Return to the OpenCode chat and ask:

Explain demo.ts in three bullets. Do not modify anything.

The model should answer in the normal OpenCode interface. /models is the safe place to confirm the provider and selection without reopening a credential prompt.

If Vercel AI Gateway is missing, repeat the connection flow outside the recording or shared session. If the task fails after model selection, verify that the chosen model is still available to your team.

Done-When

/models shows Vercel AI Gateway
A Gateway model is selected
OpenCode completes the read-only task
No credential prompt or key value appears in the recording or repository

Solution

OpenCode stores the provider connection through its setup flow. Use /models for later model changes, and keep /connect for private setup sessions where credential entry will not be captured.