MyContext - Alibaba's open-source Agent context infrastructure for Qianwen Office - AiBoss

What is MyContext?

MyContext is an open-source Agent context infrastructure developed by Alibaba Cloud's Qianwen Office team. Running primarily locally, it processes multi-source work data from platforms like DingTalk and Lark (IM communications, documents, meetings, etc.) into continuously updated, traceable, and personalized "work archives," containing responsibilities, collaboration relationships, and work conclusions. When encountering conflicting information, semantic analysis is used to determine whether to merge or overwrite; otherwise, user confirmation is required. Based on these archives, dedicated Agents can be built to enable work information queries and personalized automatic responses.

The main functions of MyContext

Building work filesIt consolidates multi-source data from DingTalk, Lark, IM, documents, meetings, etc., into a continuously updated personal work archive, covering responsibilities, collaboration relationships, and discussion conclusions.
Information is traceableEach piece of information in the archive is marked with its source, time, and content, and can be traced back to its original source.
Conflict Information Processing: Use semantic analysis to determine whether to merge or cover contradictory information. If a decision cannot be made, the user must confirm the decision and the information will no longer be covered by the model.
Search Q&AIt supports natural language query of work history and provides answers with sources by combining full-text search, semantic recall and graph query.
Digital cloneBased on profiles, a dedicated Agent can be built, which can draft and send work information in the user's tone, and operations such as sending require user authorization.

How to use MyContext

Download and Deploy: Get the project from the GitHub repository and run the desktop application on your local device in a local-first manner.
Authorized access data sourceConnect to and authorize IM systems such as DingTalk and Lark, as well as work data sources such as documents and meetings.
Import dataImport your personal work data into MyContext, and the system will automatically complete the standardization process and incremental synchronization.
Generate working filesThe system automatically generates and continuously updates exclusive work files that include responsibilities, collaborative relationships, behavioral habits, and discussion conclusions.
Build a dedicated agentGenerate a personalized Agent assistant based on your profile, which can be used to query work information, summarize knowledge, and reply to messages in your own voice.
Manual confirmation of key stepsWhen encountering contradictory information or sending messages that could have consequences, the user must confirm before execution to ensure accuracy and security.

The core advantages of MyContext

Contextual infrastructure locationIt processes scattered work data in IM, documents, and meetings into exclusive work files that the Agent can understand, thus completing the memory base for the Agent's work scenarios.
Intelligent governance of conflict informationSemantic analysis is used to determine whether information should be merged or covered. If uncertain, the user confirms the information and the model will not cover it again, thus avoiding information loss in traditional solutions.
The entire process can be traced back to the anti-hallucinationEach piece of information is marked with its source, time, and content, making the answers verifiable and improving the credibility of the results.
Local privacy is prioritized.Data and indexes are stored on the user's local machine; AI is merely a user, not the owner, thus aligning with enterprise data security requirements.
Human-machine safety gate mechanismOperations with consequences, such as sending messages, must be explicitly authorized by the user, balancing automation with security and controllability.

MyContext project address

Project official website：https://mycontext.md/
GitHub repository：https://github.com/openTrinity/mycontext

Comparison of MyContext with similar products

Comparison Dimensions	MyContext (Alibaba Qianwen Office)	Mem0
Publisher	Alibaba Cloud · Qianwen Office Team	Mem0 AI (US startup, founded in 2023)
Product Positioning	Agent context infrastructure for office scenarios	General Agent Memory Layer (Memory as a Service)
Product Form	Native-first desktop applications + nine-layer complete architecture	API/SDK middleware, which needs to be integrated by developers.
Target users	For individuals and businesses in the workplace, it's ready to use right out of the box.	AI application developers
Data source	Natively integrates with DingTalk, Lark, and other IM, document, and meeting recording services.	It primarily handles dialogue history; the data source needs to be integrated independently.
Memory architecture	Memory graph + Full-text/semantic/graph hybrid search	Vector-based memory retrieval and extraction
Conflict Information Processing	Semantic analysis determines merging/coverage; when uncertain, manual confirmation is required, and the model cannot cover the result.	Heuristic add-or-update mechanisms, without human adjudication, may allow old facts to "resurrect."
Retrospectivity	Each piece of information is marked with its source, time, and content, making it fully traceable.	The traceability of memories is weak, and there is a problem of "referring to memories but being unable to pinpoint the source".
Privacy and security	Local priority, data and indexes are stored on the user's local machine.	Cloud hosting is the default; self-hosting requires additional deployment.

Application scenarios of MyContext

Smart message replyAgent automatically drafts and replies to DingTalk/Lark work messages based on the user's work profile and communication habits, using the user's tone of voice.
Job Information Q&AUse natural language to query an individual's work history (e.g., "What was the plan decided in last week's meeting?"), and include the original source of the answer.
Personal knowledge accumulationAutomatically organizes the responsibilities, collaboration relationships, and discussion conclusions scattered in IM, documents, and meetings into a continuously updated work archive.
Collaboration Relationship Analysis: By presenting the connections between people, projects, and topics through memory maps, the context of cross-departmental collaboration can be quickly clarified.
New employee/job transfer handoverComplete and traceable work records reduce the costs of job handover and knowledge transfer.