First run
Open a workspace, wait for the engine to start, and sign in with your ML.ai access token.
Open a project folder
ML.ai requires an open workspace folder. It does not operate on loose files opened outside a folder. Open the project you want to work in before starting ML.ai.
Start the engine
Click the ML.ai icon in the Activity Bar. You'll see a startup screen with a restart control while the bundled engine comes up. This is normal: the engine is a local process that ML.ai launches and manages on your behalf, and there's no external server to reach.
Starting the engine…
A local process ML.ai launches and manages for you. No external server to reach.
Restart engineIf the engine never becomes ready, see troubleshooting.
Sign in with your ML.ai access token
ML.ai needs an ML.ai access token before it can send a single message. There's no bundled trial access and no separate model provider key to collect, so this step is required for every install.
Open Provider settings
Open ML.ai Settings → Provider.
Paste your access token
Paste your ML.ai access token (it looks like ml-ai_ followed by 32 characters).
Sign in
Click Sign in. The token is written to VS Code's SecretStorage. It is never shown again after saving, never returned to the panel, and never included in copied diagnostics.
ML.ai does not migrate or reuse an existing OpenCode CLI credential, even if one is already configured on the same machine. Every install starts signed out.
Next: first requests.