Qwen Open-Sources MyContext, Ushering in the Era of Enterprise Context

Building a New Context Infrastructure for Agents

Recently, Qwen Office officially open-sourced a new context infrastructure called "MyContext."

It processes fragmented and diverse personal work data into dedicated work files that are easier for Agents to understand, enabling Agents to truly comprehend users and real-world business workflows. This allows efficient human-AI collaboration during decision-making and execution, significantly enhancing performance in office scenarios.

Within just one week of launch, the project has already garnered over 1,000 stars on GitHub, with its popularity steadily rising.

Enabling Agents to Truly Understand Users and Real Business Workflows

Why Context Infrastructure Is Needed

As model capabilities rapidly improve and Agent Harness architectures evolve, general-purpose Agents are becoming increasingly mature, significantly boosting individual productivity. However, due to the lack of contextual awareness from real working environments, these Agents struggle to integrate into actual business processes, limiting organizational productivity gains. According to MIT's NANDA report, approximately 95% of enterprise-level generative AI pilot projects fail to generate returns, primarily due to insufficient data infrastructure, which prevents AI systems from integrating into existing workflows.

A Continuously Updated Work Profile

Unlike traditional knowledge bases or vector databases, MyContext runs locally on users' devices and automatically aggregates multi-source data from instant messaging, documents, meeting notes, collaboration platforms, and both online and offline channels, consolidating them into a continuously updated personal work profile. This profile includes job responsibilities, collaboration networks, behavioral patterns, and conclusions from work discussions, allowing Agents to quickly query and reason during task execution. Each piece of data can be traced back to its original source, clearly labeled with origin, timestamp, and content, ensuring transparency and helping mitigate AI hallucinations, thereby greatly improving task accuracy.

The Technical Innovation Behind MyContext

Qwen Office employs innovative algorithms and engineering designs when handling critical tasks such as time-series and conflicting data. For example, regarding contradictory information in instant messages, traditional approaches typically adopt either "use the latest" or "use the highest confidence" methods, both of which can lead to information loss in Agent scenarios. When MyContext detects conflicts, it first performs semantic analysis to determine whether the data should be merged or overwritten. If automatic resolution isn't possible, it prompts the user for confirmation. Once confirmed by the user, the model cannot override this data again, thereby greatly ensuring data validity and consistency.

The Era of Enterprise Context Is Approaching

Currently, MyContext supports processing data from personal DingTalk, Feishu, and similar IM systems. After importing their data, users can build personalized Agent assistants capable of summarizing their knowledge and automatically replying to work messages in their own tone. In the future, the team plans to release an enterprise version of the context infrastructure, offering full support for mainstream enterprise systems like Salesforce and SAP, as well as local workflows, driving the arrival of the enterprise context era and accelerating overall organizational productivity.

We dedicated to serving clients with professional DingTalk solutions. If you'd like to learn more about DingTalk platform applications, feel free to contact our online customer service or email at This email address is being protected from spambots. You need JavaScript enabled to view it.. With a skilled development and operations team and extensive market experience, we’re ready to deliver expert DingTalk services and solutions tailored to your needs!