ActionWright documentation

Instead of only answering questions, ActionWright can help move a task forward. You describe what you want, choose the right context, and watch the assistant plan, ask follow-up questions, request approvals, and report results.

ActionWright is useful when a task crosses apps, files, and websites: researching a topic, organizing a folder, drafting from notes, filling out a browser workflow, or using voice when your hands are busy.

• Install or open the ActionWright Windows app.
• Add an AI provider in settings and validate the key before saving.
• Choose voice, push-to-talk, or the task window for your first request.
• For file tasks, choose the folder ActionWright should work inside.
• Start with a low-risk task such as summarizing a document or finding a file.

You can use voice for short commands, dictation, quick research, and hands-free follow-up questions. After a voice interaction starts, ActionWright can continue the conversation without requiring you to repeat the wake phrase every time.

If the assistant is speaking or working and you need to change direction, you can interrupt, stop the current response, or start a new conversation.

• Use push-to-talk when you want deliberate control over when the assistant listens.
• Use wake-word interaction when you want a more conversational flow.
• Use spoken responses when you want the result read aloud.
• Use stop or interruption when the assistant is going in the wrong direction.

For long-running tasks, ActionWright shows what it is doing and pauses when it needs your input. You can answer clarifying questions, approve sensitive actions, extend a step budget, or stop the task.

The task window can be used for both browser and file work. For file tasks, you choose the root folder ActionWright can inspect or modify. For browser tasks, you can enable browser mode and let the assistant work through a visible browsing session.

You can ask ActionWright to research a topic, collect information from pages, compare options, navigate a multi-step flow, or help with a form. When a page requires login, CAPTCHA, two-factor authentication, or other personal input, ActionWright can pause so you can take over.

Browser actions that submit data, upload or download files, change an account, make a purchase, or affect an outside system are treated carefully and may require confirmation before they continue.

• Search the web and summarize findings.
• Read pages and extract useful information.
• Navigate websites with visible progress.
• Fill forms with user approval before sensitive submission.
• Pause for login, CAPTCHA, 2FA, payment, or personal account steps.
• Let the user take control and then continue when ready.

File assistance is designed for practical local work: finding a missing document, cleaning a downloads folder, comparing files, drafting from notes, or organizing a project folder.

ActionWright can inspect common document formats such as text files, PDFs, Word documents, spreadsheets, and archive contents when supported. It asks for a folder when a file task does not yet have an approved scope.

Computer control is useful for repetitive UI work, desktop app workflows, or tasks where the needed information is visible on screen. ActionWright can look at the screen, choose the next step, and ask for confirmation around high-impact actions.

Sensitive fields, account changes, destructive confirmations, payments, and other high-risk steps should remain under user control or explicit approval.

• Open or focus apps and windows.
• Click buttons and menus when the task requires it.
• Type, paste, scroll, drag, and use keyboard shortcuts.
• Pause when a step involves private data or irreversible changes.
• Hand control back to the user when manual input is required.

Different providers support different tasks. Some can power text tasks, some can support speech-to-text or text-to-speech, and some may need a separate voice provider for voice features.

When you configure a provider, ActionWright checks whether the setup can be used and reports setup issues in practical categories such as authentication, quota, model availability, rate limits, or network connectivity.

• You can stop a task while it is running.
• You can deny or revise a proposed action.
• You can take over browser or desktop steps when needed.
• ActionWright asks for context when a task is ambiguous.
• Private and sensitive fields are handled carefully rather than hidden behind automation.