MaxCompute Agentic Ecosystem Debut: Let AI Directly "Read" Your Data Warehouse
As large language models and agent technology sweep across every industry, the data analytics field is undergoing a quiet revolution. MaxCompute is officially launching its Agentic toolkit suite, covering the complete chain from desktop interaction to AI automation — making data analysis as easy as chatting, and making AI Agent the hardest-working data engineer on your team.
What Do Data Platforms Need in the Agentic Era?
In the past, an analyst's day often looked like this: open a SQL IDE to write queries, switch to the data catalog to check table schemas, open a visualization tool to build charts, and finally export an Excel file to send to the business team. The tool chain was long, context switching was frequent, and a lot of time was spent "finding data" and "tidying data" rather than generating real insights.
As AI Agents begin to penetrate office scenarios, a natural question arises: Can a data platform also become a "native environment" for Agents?
MaxCompute's answer is: yes, and much more.
Today, we officially release the Agentic toolkit suite. Five core components, spanning from the desktop to the server, from human-agent interaction to machine autonomy, build a complete data-agent infrastructure:
- AI Data Exploration Client — an integrated desktop data workbench
- MCP Server (MCMCP) — the data service access layer for AI Agents
- MaxCompute Skills — semantic skill packs for Agents
- CatalogAPI SDK — RESTful interfaces for data catalog management
- MaxCompute CLI — command-line tool for scripting and Agent integration
AI Data Exploration Client: One Tool for the Full Data Development Chain
Imagine this scenario: you say to your computer, "Show me the order trend over the last seven days," and a few seconds later a high-quality chart with a trend line appears in front of you — no SQL writing, no tool switching.
That is the experience the AI Data Exploration Client aims to deliver.
First of all, it is a professional SQL IDE: syntax highlighting, intelligent completion, query history, and multi-file management — everything data developers need on a daily basis. But it goes further: 12 chart types and a pivot tableturn query results into visualizations in seconds; AI Data Exploration supports both chat mode and agent mode, driving data analysis with natural language; CSV/Excel/Parquet import/export and WebDAV mount remove format and platform restrictions on data flow.
It supports both macOS (Apple Silicon) and Windows, so you get a consistent experience no matter which device you use. Client v1.2.9 already supports single-file and batch download of Blob files to a specified directory, making data assets easily accessible. PPT, Excel, and other files can be opened with local applications in one click, enabling seamless local workflow integration.

MCP Server: Let Agent Directly "Read" Your Data Warehouse
If the AI Data Exploration Client is the desktop entry point for "human + Agent," then MCP Server (MCMCP) is the "standard interface" for Agents entering the MaxCompute world.
Based on the Model Context Protocol (MCP), MCMCP enables various AI Agents to directly perform large-scale data analysis and intelligent operations. Agents can browse data catalogs, search metadata, and execute read-only SQL queries just like humans — and the server enforces read-only protection on the server side, so data security has a baseline.
It has already been integrated with mainstream Agent clients such as OpenClaw, DataWorks Agent, Qwen Code, QoderWork, and MaxAgent, and supports multiple authentication methods including AK/SK, STS, Credentials URI, and ECS RAM Role, making enterprise deployment hassle-free.
MaxCompute Skills: Equip Agents with Professional Skill Packs
Having an interface is not enough; an Agent also needs to "understand the business." MaxCompute Skills are a set of semantic capability packages for Agents, equivalent to giving an Agent a "MaxCompute professional engineer certificate."
The current Information Schema semantic pack covers system-level operations scenarios such as storage pressure diagnosis, cost analysis, task failure analysis, permission auditing, and hot table observation. The Agent no longer needs users to teach it step by step how to query metadata or read execution plans; it knows what to ask and what to look at.
Used together with MC MCP, the Agent evolves from "being able to connect to the database" to "being able to troubleshoot problems independently."
CatalogAPI SDK: Manage Your Data Assets with Code
For developers who need deep customization and integration, CatalogAPI SDK provides the complete RESTful interface for MaxCompute data catalog management (Java/Python dual-language support).
CRUD operations on data tables, external data source connection configuration, fine-grained Role/Policy permission control, Taxonomy policy tags for column-level permissions, DataPolicy dynamic data masking, and even multi-version management of machine learning models — every aspect of data governance can be orchestrated with code.
MaxCompute CLI: The Data Steward in the Command Line
For scripting automation and CI/CD scenarios, MaxCompute CLI (maxc) provides the lightest-weight access method.
It is distributed through the Alibaba Cloud CLI, making installation and updates seamless. Structured JSON outputmakes programmatic parsing effortless, with full coverage on Linux/macOS/Windows. Whether it is scheduling data jobs in cron tasks or letting Agents perform batch operations through the command line, it is a reliable infrastructure.
From "Tools" to "Ecosystem"
Individually, each component solves a specific scenario problem; together, they form a complete Agentic data infrastructure:
- Humans interact with data naturally through the AI Data Exploration Client.
- Agents autonomously complete analysis and operations through MCP Server + Skills.
- Developers deeply integrate and automate through CatalogAPI SDK and CLI.
This is not a simple pile-up of product features, but an architectural upgrade for MaxCompute in the AI era — from "data warehouse" to "data intelligence platform." The value of a data platform lies not only in its own capabilities, but also in its deep integration with daily collaboration tools. MaxCompute is actively connecting with mainstream office ecosystems such as Feishu and DingTalk, making data flow boundaryless.
It supports batch import of Feishu cloud disk files (PDF/DOC/PPT/IMG) into MC multimodal tables for intelligent text extraction, one-click ingestion of Feishu sheets/multidimensional tables for deep computing and analysis, and one-click sharing of MC analysis results back to Feishu sheets — from unstructured data mining to collaborative distribution, the whole process is seamlessly connected and efficiency is improved several times over.
Get Started Now
- AI Data Exploration Client: see the MaxCompute AI Data Exploration Client documentation
- MCP Server and Skills: see the MCP Overview and Access documentation
- CatalogAPI SDK: see the CatalogAPI SDK User Guide
- MaxCompute CLI: see the MaxCompute CLI documentation
If you have any questions, please contact us through a support ticket.
Read previous post:
Read next post:
-
-
Alibaba Cloud Big Data and AI
31 posts | 0 followers
Follow
Comments
-
-
Alibaba Cloud Big Data and AI
31 posts | 0 followers
Follow
Related Products
-
Big Data Consulting for Data Technology Solution
Alibaba Cloud provides big data consulting services to help enterprises leverage advanced data technology.
Learn More -
Big Data Consulting Services for Retail Solution
Alibaba Cloud experts provide retailers with a lightweight and customized big data consulting service to help you assess your big data maturity and plan your big data journey.
Learn More -
MaxCompute
Conduct large-scale data warehousing with MaxCompute
Learn More -
ApsaraDB for HBase
ApsaraDB for HBase is a NoSQL database engine that is highly optimized and 100% compatible with the community edition of HBase.
Learn More
More Posts by Alibaba Cloud Big Data and AI
- From Common Crawl to Training-Ready Corpora: Cleaning, Annotation & Vectorization with EMR Serverless Daft
- One Spark SQL, Multiple Catalogs: Unified Data Processing with EMR Serverless Spark
- Checkpoint-Based Resumption in EMR Serverless Ray Jobs
- RAG Architecture Beyond Vector Search | Hologres
- Alibaba Cloud Hologres Tops the Official TPC-H 3,000GB Benchmark, with the Best Price/Performance
- Big Data & AI Platform Monthly Newsletter—July 2026
- Big Data & AI Platform Monthly Newsletter—June 2026
- Alibaba Cloud EMR Serverless StarRocks Lakehouse Multimodal Retrieval: One SQL for Full-Text, Scalar, and Vector Hybrid Search
- Alibaba Cloud EMR Serverless Spark AI Function Multimodal & Autonomous Driving Practice