Native browser control
Understand page structure, click, fill, upload, and continue across pages—then read the resulting state.
See how it works →UQTBrowser · Built on Chromium
A browser that finishes the task, not just the answer
UQTBrowser goes beyond chat. With your permission, it can understand and operate web pages, work with Windows apps, files, and software, write and run scripts, and carry multi-step tasks through to a verifiable result.
- Windows 10 / 11 · x64 / ARM64
- 中文 + English
- Built on Chromium
- Version 19.25.211
- Updated 2026-08-20
Auto mode · Permission aware
Review this page, create an action list, and save it locally.
Four steps planned and executed
1Read and understand the pageDone
2Extract decisions and risksDone
3Create an organizing scriptDone
4Save and verify the outputChecking
Describe the next goal ↑
Native web controlRead · Click · Fill · Verify
Windows assistanceFiles · Apps · Commands · Scripts
General-purpose AI agentFrom response to execution
3 browser modesAuto · Headless · Real
Multi-protocol downloadsLinks · BT · Magnet · ED2K
51 PDF toolsConvert · OCR · Batch
General-purpose AI agent
More than chat. An AI agent that can act.
From web workflows to Windows maintenance and scripting, UQTBrowser turns “tell me how” into “help me get it done”—under your control.
Windows assistance
Work with files, software, commands, and standard controls in desktop applications.
Script creation and execution
Write code for a goal, save it, run it, and inspect the real output instead of stopping at a code block.
Multi-step completion
Plan and verify each step, while pausing for CAPTCHAs, sign-in checks, or sensitive approval.
One workflow, not one answer
01
Understand the environmentPage, files, desktop, and goal
Observe02
Select the right toolWeb control, command, script, or UI
Act03
Hand off human checkpointsCAPTCHA, sign-in, and sensitive approval
Pause04
Verify the real outcomeRe-read the page and actual output
CheckNative integration
From the web to Windows, without losing the task
The agent covers a broad range of digital work that can be carried out through web pages, files, the Windows command line, or desktop interfaces. It remains subject to your authorization, current OS permissions, website rules, and security boundaries.
- Auto, headless, and real-browser modesUse an isolated background environment or the page you can see.
- Human handoff for verificationIt does not solve or bypass CAPTCHAs and identity checks.
- Permission gates for sensitive actionsChoose Read-only, Standard, Ask for high risk, or Full automation.
Built-in download center
Direct links and peer-to-peer tasks in one place
Manage HTTP, HTTPS, FTP, BitTorrent files, magnet links, and ED2K tasks, with pause and resume, retry, task recovery, and file hash details.
- Up to 32 threads on supported direct downloadsFor HTTP, HTTPS, and FTP tasks; actual speed depends on the source server and your network.
- Peer-to-peer adapts to source healthBitTorrent, magnet, and ED2K use available peers rather than a fixed thread count.
- Hash information is visibleInspect MD5, SHA-256, and CRC32 for supported tasks.
UQT Download CenterTask recovery on
HTTPHTTPSFTPBTMAGNETED2K
Connection note: Up to 32 threads applies to supported direct-download tasks; server limits, network conditions, and source availability affect real speed.
PDF & offline OCR
Document tools built into the browser
Use 51 PDF tool entry points for common conversion, page operations, batch workflows, and offline Chinese and English OCR. Always review critical documents.
Convert PDF to DOCX, XLSX, Markdown, or images, and combine images into a PDF.
Reorganize pages, add watermarks, and compress documents without installing a separate suite.
Recognize Chinese and English text locally and add a searchable text layer to scanned PDFs.
Powerful by design. Controlled by you.
Broad capability is not unlimited permission. The AI runs within your authorization and current Windows privileges, and pauses for sign-in verification, CAPTCHAs, and explicitly sensitive actions.
Four permission levelsRead-only, Standard, Ask for high risk, and Full automation.
Execution isolationHeadless mode cannot route around its boundary into the real desktop.
Clear data choicesLocal tools and cloud AI processing are explained separately in the Privacy Policy.
FAQ
What you should know before you start
Capabilities, pricing, data, and release status—without hiding the important constraints.
How is UQTBrowser different from an AI chat app?
The AI is connected to native browser and Windows tools. With permission, it can operate pages, work with files and apps, run scripts, and continue through multi-step work instead of only describing the steps.
Can the AI change my computer without permission?
The default Standard level keeps mutation and sensitive actions behind permission gates. Commands, software installation, irreversible cleanup, and other high-risk actions may still require individual approval. Read-only mode is also available.
Which download protocols are supported?
HTTP, HTTPS, FTP, BitTorrent files, magnet links, and ED2K. Supported direct-download tasks can use up to 32 threads; peer-to-peer speed depends on available sources.
Are PDF and OCR tasks local?
The listed offline OCR capability can run on your device. If you ask the AI to analyze a document, screenshot, or attachment, the task content may be sent to the AI service you select or configure.
Is UQTBrowser free?
The browser core is free. The official AI service is billed separately and requires AI Pro or available tokens; current pricing and allowances are shown in the client.
Can I download UQTBrowser for Windows today?
Yes. Version 19.25.211 is available as separate native installers for Windows 10 and 11 x64 (Intel / AMD) and ARM64, served from this site's first-party browser.uqitong.com download URLs.
Turn a browser session into a finished workflow
Explore what the AI agent can do, or download the current public Windows build directly from our first-party download server.