SetupAI
繁
SourceStalecollected in 18m

ima Copilot fully released with new Knowledge Agent skills

Read original on 36氪
#agentic-ai#knowledge-management#productivity

A new platform for building and sharing Knowledge Agents that integrate directly with personal data.

30-Second TL;DR

What Changed

Copilot feature is now accessible to all users without a waitlist.

Why It Matters

The ability to build and share custom Skills on top of personal knowledge bases significantly lowers the barrier for creating specialized AI agents.

What To Do Next

Explore the Knowledge Plaza and try building a custom Skill using your own documentation to automate your workflow.

Who should care:Developers & AI Engineers

Key Points

  • •Copilot feature is now accessible to all users without a waitlist.
  • •New 'Knowledge Plaza' allows users to publish and discover custom AI Skills.
  • •Users can now call upon their stored documents and knowledge during Copilot task execution.
  • •Initial Skills include integrations with WeChat Reading and Tencent Recruitment.

Deep Insight

Background and context from public sources — not the original article. 21 sources cited.

Enhanced Key Takeaways

  • •ima Copilot is powered by Tencent's proprietary Hunyuan large language model and incorporates multimodal AI capabilities, allowing for tasks like image generation from text prompts and multimodal search.
  • •The Copilot features a built-in memory system that retains user context, habits, and ongoing tasks across different scenarios, aiming to reduce repetitive input and provide continuous assistance.
  • •Beyond general document storage, users are allocated 1GB of dedicated storage to upload and manage various file formats, including Word, PDF, PNG, and JPG, for building their personalized knowledge bases.
  • •The integration with WeChat Reading, offered as a custom Skill, enables AI assistants to directly access and analyze users' private bookshelves, highlighted notes, and reading habits to deliver highly personalized book recommendations and insights.
  • •The platform is designed for broad accessibility, with clients available across multiple operating systems, including Mac, Windows, iOS, Android, and HarmonyOS.

Competitor Analysis

Core Functionality
ima Copilot (Tencent)
AI-powered workspace with customizable Knowledge Agent skills, personal knowledge base, multimodal AI, intelligent writing, research assistance.
Microsoft Copilot Studio
Low-code platform for building custom AI agents (copilots) within Microsoft ecosystem.
Google Gemini Enterprise
Multimodal AI with native Google Workspace integration, Deep Research, real-time web access.
Dust
No-code AI agent platform with cross-platform data connections and LLM-native actions.
Customization/Skills
ima Copilot (Tencent)
'Knowledge Plaza' for discovering/publishing custom AI Skills; integrates with WeChat ecosystem.
Microsoft Copilot Studio
Uses Power Platform connectors to integrate with business systems; custom copilots for Microsoft 365 surfaces.
Google Gemini Enterprise
Simpler visual agent designer; agents process documents, images, videos.
Dust
True no-code agent building; connects to 100+ tools; unlimited knowledge sources per agent.
Pricing (approx.)
ima Copilot (Tencent)
Not explicitly detailed for full release; free Mac client mentioned previously.
Microsoft Copilot Studio
$200/month for 25,000 messages/month (Copilot Studio); $30/user/month (Microsoft 365 Copilot Enterprise); $20/month (Copilot Pro).
Google Gemini Enterprise
$20/month (Gemini Advanced, included with Google One AI Premium).
Dust
$29/user/month.
Key Integrations
ima Copilot (Tencent)
WeChat Reading, Tencent Recruitment, Tencent ecosystem.
Microsoft Copilot Studio
Microsoft 365 (Word, Excel, PowerPoint, Outlook, Teams), Power Platform.
Google Gemini Enterprise
Google Workspace (Gmail, Docs, Sheets, Calendar, Drive), Microsoft 365 (SharePoint, Outlook).
Dust
Google Drive, Notion, GitHub, Slack, Confluence, Salesforce, 100+ tools.
Underlying Model
ima Copilot (Tencent)
Tencent's Hunyuan large language model and Mixed Big Model technology.
Microsoft Copilot Studio
OpenAI's GPT (currently 4-turbo, hosted by Microsoft).
Google Gemini Enterprise
Google's multimodal AI.
Dust
LLM-native actions.

Technical Deep Dive

  • Foundation Model: ima Copilot is built on Tencent's proprietary Hunyuan large language model (LLM) and leverages its 'Mixed Big Model technology'.
  • Multimodal AI: It integrates multimodal functionalities, enabling tasks such as generating images from text prompts and performing multimodal searches where users can query information from uploaded images or screenshots.
  • Knowledge Agent Architecture: The system supports a 'skills ecosystem' with built-in official skill packages and the ability to load extended skills on demand, suggesting a modular approach to agent capabilities.
  • Personalized Knowledge Base (RAG-like): Users can construct a bespoke knowledge base by importing local files or integrating web content, allowing the AI to generate responses based specifically on this curated content, indicative of a Retrieval-Augmented Generation (RAG) architecture.
  • Memory System: Features a built-in memory system with modules for Copilot settings, user profiles, long-term memory, and expertise techniques to maintain context across interactions.
  • Natural Language Processing (NLP): Integrates advanced NLP for streamlining knowledge acquisition, research, and creative tasks.

Future ImplicationsAI analysis grounded in cited sources

The 'Knowledge Plaza' will foster a robust ecosystem of specialized AI agents.
By allowing users to publish and discover custom Skills, ima is creating a marketplace that encourages developers and users to build highly niche and effective AI tools for specific tasks and knowledge domains.
Personalized AI assistants will become increasingly sophisticated and integral to individual workflows.
The ability to deeply integrate personal knowledge bases and leverage a built-in memory system means ima Copilot can evolve into a highly tailored and context-aware assistant, significantly enhancing personal productivity and learning.
Tencent will further leverage its extensive ecosystem to differentiate its AI offerings.
The existing deep integration with platforms like WeChat Reading suggests a strategy to embed ima Copilot within Tencent's vast user base and services, creating a unique value proposition difficult for competitors to replicate in the Chinese market.

Timeline

2016-04
Tencent AI Lab established, focusing on AI research and development.
2025-04
Tencent announced a major hiring initiative, aiming to recruit 28,000 interns over three years, with a strong emphasis on AI and technical roles.
2025-07
ima Copilot was positioned as an AI-driven intelligent work platform by Tencent, offering 1GB storage and the ability to build specialized knowledge bases.
2026-04
Tencent IMA officially launched Knowledge Agent - Copilot, initially available by application only, featuring a built-in memory system for personalized agent creation.
2026-05
WeChat Reading introduced an AI feature (Skill) allowing direct integration with AI assistants via an API key for personalized access to user reading data.
2026-05
ima Copilot fully released to all users, removing the waitlist and introducing 'Knowledge Plaza' with new Knowledge Agent skills.
Read original article on 36氪

Weekly AI Recap

Read this week's curated digest of top AI events →

PostLinkedIn

AI-curated news aggregator. All content rights belong to original publishers.
Original source: 36氪 ↗

This is a summary, not the original. Read the source, or get the weekly briefing.

The weekly digest

One email a week. Unsubscribe anytime.