Next message suggestions
After each Claude or Codex reply, Vato suggests the message you are most likely to send next, greyed out in the empty composer.
What it does
When a Claude or Codex turn finishes, Vato asks a light model to predict the single message you are most likely to type next, based on the visible conversation. The suggestion appears greyed out in the empty message field. It is never sent on its own: you accept it, edit it, or ignore it.
- Claude conversations use Claude Haiku 5.5; Codex conversations use GPT-6 Luna.
- The request runs on the same account as the conversation, so it uses a small amount of that account's usage after each reply.
- Other providers (Cursor, OpenRouter, ACP…), Bots and Design conversations do not get suggestions.
Use a suggestion
| Where | How |
|---|---|
| Desktop and Remote | Press Tab in the empty message field. Typing the beginning of the suggestion keeps the rest visible; Tab completes it. |
| Web companion | Press Tab in the empty message field. |
| Mobile | Tap Use suggestion above the message field. |
The text is inserted into the field so you can still change it before pressing Enter. The suggestion disappears as soon as you send a message, and while the agent is working.
Where the instructions come from
If you are signed in to Codex on this computer, Vato reads the prediction instructions published for your ChatGPT account (the same ones the Codex app uses), at most once a day. Otherwise, or if that request fails, Vato uses its own built-in instructions. Only the instructions are kept; your login is not stored by this feature.
Privacy
Suggestions are computed on the computer that runs the conversation and are shared only with your own devices (other computers in Remote, the web companion and your phone). Team members who can see a conversation never receive your suggestions.
Turn it off
Open Settings → General and switch off Suggest my next message. Existing suggestions disappear immediately and no more requests are made.