FAQ

Drag and drop the file anywhere onto the extension side panel — a "Drop files or images here" target appears (images and documents up to 10MB) — or click the + button next to the prompt box and choose Add Files. The same + menu also has Add Sheets (attach a Google Sheet whose rows drive a workflow) and Add Tabs (select browser tabs for multi-page workflows). The cloud dashboard composer accepts attachments the same way.

Attach the file first, then reference it in your prompt — for example "fill this job application and upload my resume." The agent can read attached documents and upload any attached file into a page's file input while filling forms. For security it cannot browse files on your computer that you haven't explicitly attached to the task.

In the extension, downloads go through Chrome's normal download flow, so they're in your browser's Downloads folder — open chrome://downloads in a new tab to see every file the agent saved with a "Show in folder" link (the run's result also lists each file with a 💾 chip that opens that page). In cloud runs the files are backed up to your Files and attached to the run. If the agent reported several files but you only find one, Chrome blocked the extra automatic downloads: click the blocked-downloads icon at the right end of the address bar, choose "Always allow", and re-run. See Files & PDFs.

Every account starts with free monthly credits that work across the extension, cloud runs, and the API. Bringing your own LLM API key shifts model usage to your own provider instead of platform credits, and nothing deducts from the free plan's daily Free Mode allowance. The free plan is sponsored either way (a small card appears beside results); ad-free runs come with paid plans. Current plans, credit allowances, and prices are always at Pricing; Rover site subscriptions are priced separately at Rover Pricing.

That is the per-task credit ceiling: a cap on what a single run may spend before it checks with you, so a task that loops or crawls further than expected can't quietly burn your balance. The default is 100 credits on paid plans and 25 on Free. Answer "Continue" to resume from exactly where it stopped with a fresh budget (nothing already done is repeated), or "Stop here" to keep the result and end the task.

If your work routinely costs more than that, raise the ceiling once instead of answering every time: in the extension it is Settings → Agent Settings → "Credit Ceiling / Task", and in the cloud it is Settings → Storage & Execution. 0 turns the check off. Full details in Web Agent → Credit Ceiling per Task.

Open Model Settings in the composer and use the "Add your LLM API" flow — it accepts keys from any major provider (OpenAI GPT models, Google Gemini, DeepSeek, OpenRouter models, or a local/custom endpoint), and you can also sign in with your ChatGPT plan directly. For Gemini, typing /add-gemini-key in the extension walks you through getting a free key from Google AI Studio. Keys are stored encrypted and work in cloud runs too. See Getting Started → Models & Bring Your Own Key.

Managed runs use GLM 5.3 Flash on the free plan (in ad-supported Free Mode, with a daily fair-use limit) — DeepSeek Flash, DeepSeek Pro and Gemini Flash selectable on paid plans. Gemini Pro has been retired as a managed option; saved tasks that request it run on Gemini Flash automatically. With BYOK you can run any model your provider offers. The Changelog always lists the latest model updates.

Voice input works: tap the microphone button in the composer to speak your task, and once WhatsApp is connected you can send voice messages to trigger workflows (see Cloud & Mobile). However, the agent cannot listen to audio playing inside a web page — it reads the page's structure (DOM), so tasks that depend on hearing page audio (e.g. transcribing an embedded player) are not supported.

A cloud run starts on about:blank while a fresh browser instance boots and the automation script loads. As soon as the workflow performs its first navigation, the live preview switches to the real page.

Extension schedules run inside your Chrome with your logged-in sessions, so they execute while the browser is open — a missed run fires automatically when Chrome reopens. For true 24/7 execution, export the workflow to Cloud and schedule it there. See Schedules.

In the extension, yes — it works in your browser with your existing sessions, no credential sharing needed. Cloud browsers start fresh, but Cookie Sync securely uploads your sessions so cloud runs can access authenticated sites too.

If the error appears before a draft tool opens, refresh your rtrvr sign-in and retry once. If the tool was created and fails when it starts, sign in to the target site again, make a fresh recording, and regenerate the Subroutine once. If it fails again, keep the run and contact support: the generated request may be missing a current token or request ID. A Workspace shares credits, not member logins or target-site credentials. See Recordings → If a Token or ID Is Invalid for the full check and the details to send support.

No — the agent is DOM-native: it acts directly on browser tabs (clicking, typing, extracting against the page structure) and you can watch it work in real time. It never controls your mouse, keyboard, or anything outside the browser.

That's Rover — create your site in the Rover Workspace, copy your Site ID, and add one script tag to your site. See the Rover docs for installation and configuration.