Ant Group open-sources antv's MCP service: How to build a bridge between AI agents and data visualization?

1. What is antv's MCP service?

Ant Group's open source antv MCP service is a standardized tool calling platform built on the MCP protocol (Model Context Protocol), which aims to provide convenient data visualization capabilities for AI agents. Its core is to encapsulate data visualization tools (such as the AntV chart library) as "plug-ins" that can be called by AI models through a unified protocol, thereby achieving deep integration of agents with chart generation and data analysis tools.

The MCP protocol was proposed by Anthropic and is likened to the "USB-C interface in the AI ​​era" to solve the pain point of fragmented integration between AI models and external tools and data sources. Ant Group, through the open source antv MCP service, transforms its own technical accumulation in the field of data visualization (such as G2, F2 and other chart libraries) into a "toolbox" for intelligent entities, promoting the implementation of AI in scenarios such as data analysis and real-time monitoring.

2. What are the functions of antv's MCP service?

1. Zero-code access to AI toolchain
   
   Developers can connect AntV's chart generation, data mapping and other functions to the intelligent agent workflow in the form of "tools" without writing adaptation code, such as generating bar charts or heat maps through natural language instructions.
2. Multimodal data interaction
   
   It supports multiple modal inputs such as text, images, and time series data, and combines with Alibaba Cloud observability products (such as Log Service SLS and ARMS monitoring) to achieve cross-platform data linkage analysis.
3. Security and access control
   
   Integrate Alibaba Cloud AccessKey authentication and the principle of least privilege to ensure data privacy and operational compliance during tool invocation.
4. Dynamic resource expansion
   
   Supports dynamic registration of new tools through the MCP protocol, such as adding custom chart types or analysis algorithms to meet the needs of rapid business iteration.

3. The principle of antv's MCP service

The core architecture of antv MCP service is divided into three layers:

1. MCP Client
   
   Integrated in AI agents (such as Alipay applet, Cursor IDE), it is responsible for converting user instructions (such as "generate a sales trend chart for the past week") into standardized tool call requests.
2. MCP Server
   
   Deploy the AntV chart library and data processing tools, parse the parameters (such as time range, indicator type) after receiving the request, call the corresponding chart engine to generate results, and return the visual data stream through the Streamable HTTP protocol.
3. Protocol Layer
   
   The Streamable HTTP protocol based on MCP implements two-way communication, supports disconnection recovery and batch processing, and optimizes the efficiency of large-scale data interaction. For example, the agent can request multiple chart rendering tasks at the same time, and the server returns them in batches according to priority.

Technical highlights:

• Natural language to query
  Built-in SLS natural language parsing module converts fuzzy instructions (such as "most visited applications") into structured query statements.
• Dynamic metadata annotation
  Automated permission management is achieved by marking tool risk levels through Tool Annotations (such as "destructiveHint" marks deletion operations).

4. Project Address

• GitHub repository https://github.com/antvis/mcp-server-chart
  

5. Application Scenarios

1. Intelligent data analysis assistant
   
   Corporate employees can generate interactive dashboards in real time through natural language commands (such as "show Q2 revenue comparison by region"), replacing manual operations of traditional BI tools.
2. Automated operation and maintenance monitoring
   
   Combined with the log service SLS, the AI ​​operation and maintenance assistant automatically analyzes the system exception log and calls AntV to generate a fault timeline map.
3. Visualized teaching in education
   
   Teachers assign homework (such as "drawing a population growth curve") through conversational AI, and students call MCP services to complete data visualization tasks.
4. Financial risk control real-time report
   
   When the risk control system detects abnormal transactions, it automatically generates a visual report and pushes it to the decision-making end to improve response speed.

Tags :
#datavisualization #MCPprotocol #AIagent #opensourcetools #AntGroup 

Conclusion
Ant Group's open source antv MCP service marks the evolution of the AI ​​intelligent body ecosystem from "single model capability" to "tool collaboration network". With more developers joining in, innovative applications such as "generating a data screen in one sentence" and "automatically writing analysis reports by AI" may emerge in the future, pushing data intelligence into the era of popularization.