How Marvis’s Six AI Agents Replace 80% of Your PC Tasks
Tencent’s free Marvis AI assistant, launched May 2026, operates at the OS level to read and write files, adjust system settings, perform remote Android control, and build searchable knowledge bases, promising to automate up to 80% of everyday computer work.
On May 20 2026 Tencent’s App Store quietly released Marvis, a completely free OS‑level AI assistant that can directly manipulate Windows and macOS, unlike ChatGPT or Wenxin Yi which are limited to reading manually uploaded content.
Feature comparison shows Marvis supports file read/write, system‑parameter changes, full‑disk image semantic search, Android remote control, offline local‑model operation, custom scheduled automation, and multi‑agent visual task progress, whereas ordinary web/client AI tools lack these capabilities.
Core Architecture
Marvis is built from six specialized agents:
Marvis Leader : overall scheduler that decomposes complex requests and assigns subtasks.
File Agent : document classification, deduplication, format conversion, OCR extraction.
System Agent : modifies system parameters, runs hardware diagnostics, optimizes game performance, repairs faults.
Application Agent : launches and interacts with desktop software (including Windows Android APKs) by simulating clicks.
Retrieval Agent : full‑disk semantic search for files, images, and cross‑knowledge‑base comparisons.
Scheduler Agent : executes periodic tasks, data push, and automatic backups.
Technical advantages stem from deep Tencent‑Microsoft cooperation (native Windows control), Intel‑level on‑device inference (2‑10× speed boost), and a 14‑year cross‑platform ecosystem that supports Windows, macOS, Android, with iOS in development.
Installation Guide
Download the installer from marvis.qq.com (Windows 10/11, macOS 12+, Android 9+, iOS 17.6+). Minimum hardware: 6‑core CPU + 16 GB RAM + SSD for efficiency mode; 16‑core CPU + 32 GB RAM for local mode.
Log in via WeChat/QQ QR code, then grant file, system, application, and screen‑capture permissions.
Select runtime mode: “Efficiency mode” (cloud‑assisted, fast, suitable for daily writing, file sorting, ordinary office work) or “Local mode” (offline, all data stays on‑device, ideal for confidential files).
Typical Use Cases
Desktop file organization & semantic image search : batch move PDFs, locate forgotten screenshots, retrieve all 2025 contracts; AI completes in seconds versus a half‑hour manual effort.
One‑click system tweaks & hardware checks : change resolution, set high‑performance power plan, evaluate game compatibility, disable redundant DVR features.
Android remote PC control : view screen, take over desktop, install software, assist family members, even when the PC is locked.
Office automation : batch convert Word to PDF with watermarks, generate salary‑distribution charts from Excel, auto‑create PPT from reading notes, extract invoice data into reimbursement tables.
Scheduled automation : daily cleanup of download folder, weekly backup of work documents, weekly market‑price retrieval for real‑estate listings.
Knowledge‑Base Construction
Marvis can turn all local documents into a searchable knowledge base. Recommended folder hierarchy and upload workflow are provided. Supported formats include PDF, Word, Excel, PPT, TXT, Markdown; individual files ≤50 MB (larger files must be split).
KnowledgeBaseRoot
├─IndustryStandards (technical standards, regulations PDF)
├─ProjectData (ProjectA/, ProjectB/ subfolders)
├─CompetitorReports
└─PersonalNotes (notes, plans, experience summaries)After uploading, add 2‑3 keyword tags and version information to improve retrieval accuracy.
Performance Comparison
Manual search for a technical standard takes 15‑20 minutes; Marvis returns an answer in ~30 seconds. Comparing multiple documents drops from 40 minutes to 2 minutes, and compiling new‑employee training material shrinks from 1 hour to 5 minutes.
Safety and Limitations
High‑risk actions (bulk deletions, core system changes, financial transactions) require manual confirmation; AI cannot execute them autonomously.
Local mode isolates the model to prevent any cloud leakage of confidential files.
Hardware demand is high for local mode (16‑core CPU + 32 GB RAM); low‑end PCs should stay in efficiency mode.
Early versions had bugs in midnight‑crossing scheduled tasks; these have been fixed in newer releases.
Image‑only PDFs must be OCR‑processed before upload, otherwise they cannot be read.
Best‑Practice Recommendations
Use local mode for confidential contracts, payroll, and legal documents.
Use efficiency mode for everyday office work, file cleanup, and system debugging.
Always review AI‑generated content, reports, and technical conclusions before final use.
Regularly update the client to obtain new skill packs and performance optimizations.
Target Audiences
Office workers (writers, operations, R&D) – need file organization, knowledge retrieval, automated reporting.
Finance / legal professionals – require local storage, zero cloud leakage, batch contract processing.
Home users, students, seniors – need remote assistance, simple document handling, game optimization.
Marvis is free, requires no invitation code, and provides 10 million free tokens per day, making it ready for immediate use.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
AI Architecture Hub
Focused on sharing high-quality AI content and practical implementation, helping people learn with fewer missteps and become stronger through AI.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
