A little guidance.
A smoother start.
Who can use the preview?
Anyone can explore the interface, supported local conversions, and the device script reader. Paid AI calls are restricted to the owner preview. Customer account and checkout integrations require their service connections before activation.
How do I install the app?
Open the app in a supported browser. On Android, use the browser menu and choose Install app or Add to Home screen. On iPhone, use Safari’s Share menu and Add to Home Screen. This installs the web app, not a native store application.
How do I stop or retry a reply?
Tap the square Stop button while chat is generating. Interrupted text is marked as partial and is not saved to cloud history. Retry message requests a complete answer; Try another answer replaces the latest reply. Export important partial replies before leaving the conversation.
How does the task workbench work?
Open Agent workspace, describe a goal and add context. Create an AI plan, then select a task and draft its deliverable. Review and export your work. Plans and drafts are saved on this device; use New project to clear them on a shared device. The local starter template works without an AI request. External actions are not executed.
How does voice chat work?
In Voice studio, choose Voice chat and allow microphone access. Record a short question, then stop recording to send it. The owner preview transcribes it and requests an AI reply. The spoken response is AI-generated. If autoplay is blocked, use the audio player. Recording is limited to 30 seconds and 3 MB.
What are the preview limits?
The owner preview allows 20 combined chat, task-plan and task-draft requests, 3 image requests, 10 audio transcriptions, and 10 AI speech replies per day, plus two cloud agent/browser starts and one video request when its provider is configured. Voice chat also consumes a chat request. Provider credits and model access are required.
Where are my details saved?
Display names, preferences, and planning templates are saved on your device when browser storage is available. Successful signed-in chats are stored under your account. OpenAI processes prompts and audio used for AI features. The launch privacy policy and customer data controls still need completion before sales.
Can I buy a plan now?
Checkout remains closed until Stripe, verified subscription syncing and customer usage allowances are configured. Plans & billing shows the current availability and links to the payment provider when enabled.
How do I test my phone’s microphone and speaker?
Open the device audio check on the phone you want to test. Record a greeting, play it and confirm whether you heard it. The owner can also run the AI transcription/reply/speech check. Export the device report if needed. Only a person using that device can confirm audible sound.
How do autonomous runs work?
In Agent workspace, choose Autonomous sandbox. The owner can start a bounded code/file task, refresh its status, request Stop, inspect command results and download generated artifacts. Keep the page open for ten-minute cancellation checks. Closing a page does not confirm the run stopped. Cloud browser is a separate public-page mode and needs approval for each website.
What if something fails?
Keep a copy of important work, note the tool and error message, and retry only when appropriate. AI outputs can contain mistakes and should be reviewed.