AI programming assistant IDE interaction recording system based on MCP

Abstract

The invention relates to the technical field of computer information, in particular to an AI programming assistant IDE interaction recording system based on MCP, comprising: a data acquisition module configured with a standardized interaction data model including necessary fields and optional extension fields, the data acquisition module is used for receiving interaction data uploaded by an IDE (integrated development environment) client following an MCP (multi-channel protocol) through an upload action tool interface; the data storage module is used for persistently storing the interaction data received by the data acquisition module by adopting a JSON (JavaScript Object Notation) structured format; the query analysis module is used for providing a multi-mode query function and a multi-dimensional screening function through an exctdailyreport tool interface, and the query analysis module is used for organizing a query result into a structured JSON (JavaScript Object Notation) report file; and the IDE adaptation module is based on a plug-and-play mechanism of an MCP protocol, so that each IDE client can seamlessly integrate the system by configuring the MCP client, and unified acquisition, storage, tracing and analysis of interactive data in a multi-heterogeneous IDE environment are realized.

Description

Technical Field

[0001] This invention relates to the field of computer information technology, and in particular to an AI programming assistant IDE interactive recording system based on MCP. Background Technology

[0002] In the field of AI-assisted development, the interaction between developers and AI programming assistants (such as Copilot, Cursor, and Trae) generates a large amount of real-time operational data, including code editing instructions, AI response suggestions, tool call records, and file change information. This data is of great significance for evaluating the effectiveness of AI programming assistants, tracking the completion process of programming tasks, analyzing engineer productivity metrics, and optimizing AI assistant recommendation strategies.

[0003] However, existing technologies have the following limitations:

[0004] Fragmented interactive data: Existing IDEs (such as Visual Studio Code, Cursor, etc.) can only record single-dimensional data. Multi-dimensional related data such as user input, AI response, and context information are scattered and stored in different system modules, which cannot form a complete interaction link.

[0005] Lack of standardized data collection specifications: Different IDEs and AI assistants use different interactive data collection formats and methods, resulting in inconsistent data quality and making it difficult to conduct unified analysis and generate reports;

[0006] Poor IDE compatibility: Existing solutions are mostly customized for specific IDEs, and adapting to multiple heterogeneous IDEs requires writing a lot of proprietary code, resulting in high system complexity and maintenance costs;

[0007] Lack of traceability: There is a lack of a complete interaction link tracing mechanism, making it impossible to clearly trace back the execution status, time consumption, and failure reasons of complex multi-step interactions;

[0008] Insufficient query and reporting capabilities: It is difficult to perform refined queries of interaction records by multiple dimensions such as project, user, and date, and it is unable to generate structured analysis reports.

[0009] Therefore, there is an urgent need for a system that can overcome the above-mentioned technical bottlenecks and realize unified collection, standardized management, complete traceability and flexible analysis of interactive data in a multi-heterogeneous IDE environment. Summary of the Invention

[0010] To overcome the technical problems of fragmented AI programming interaction data collection, inconsistent formats, poor compatibility, insufficient traceability, and weak query and analysis capabilities in existing technologies under multi-heterogeneous IDE environments, the purpose of this invention is to provide an AI programming assistant IDE interaction recording system based on MCP. Through a standardized data model, a universal interface of the MCP protocol, complete link tracing, and flexible query and reporting tools, it realizes the unified collection, storage, traceability, and analysis of interaction data under multi-heterogeneous IDE environments.

[0011] This invention is implemented using the following scheme:

[0012] An AI programming assistant IDE interaction recording system based on MCP (Multi-Channel Programming Protocol) is disclosed. The system includes a data acquisition module, a data storage module, a query and analysis module, and an IDE adaptation module. These modules achieve data interaction and functional collaboration through the MCP protocol. Specifically, the system includes:

[0013] Data acquisition module: Configures a standardized interactive data model, which includes required fields and optional extended fields. The data acquisition module receives interactive data uploaded by IDE clients that conform to the MCP protocol through the upload_interaction tool interface.

[0014] Data storage module: Persistently stores the interactive data received by the data acquisition module using JSON structured format;

[0015] Query Analysis Module: Provides multi-mode query and multi-dimensional filtering functions through the export_daily_report tool interface. The query analysis module organizes the query results into a structured JSON report file.

[0016] IDE Adaptation Module: Based on the MCP protocol, the plug-and-play mechanism allows various IDE clients to seamlessly integrate the system by configuring the MCP client, without modifying the core system logic.

[0017] Furthermore, the required fields include user_name, project_name, ide_name, user_input, assistant_reply, and project_path, while the optional extended fields include context, tool_calls, file_changes, duration, status, and error_message.

[0018] Furthermore, the multi-mode query includes single-date query and date range query.

[0019] Furthermore, the multi-dimensional filtering includes refined filtering by user_name and project_name.

[0020] Furthermore, the upload_interaction tool interface defines a unified InputSchema input mode, eliminating the need for IDE-specific adaptation code.

[0021] Furthermore, the context field is in JSON format and records the execution context information, including file status and cursor position.

[0022] The tool_calls field is in JSON format and records a list of tools called.

[0023] The file_changes field is in JSON format and records details of file changes.

[0024] The duration field is a millisecond-level integer, recording the interaction time.

[0025] The status field is an integer, where 1 indicates successful interaction and 0 indicates failed interaction.

[0026] The error_message field is a string and provides a description of the reason for the failure only when the status field is 0.

[0027] Furthermore, the InputSchema input mode of the upload_interaction tool interface corresponds to the interactive data model, and the integrity of the required fields is strictly verified to ensure that the uploaded data structure is standardized.

[0028] Furthermore, the query parameters of the export_daily_report tool interface include date, start_date, end_date, user_name, and project_name, where:

[0029] When only the date parameter is entered, query the interaction records for that specific date;

[0030] When the start_date and end_date parameters are entered, query the interaction records between the two dates;

[0031] When paired with the user_name or project_name parameter, the query results will be filtered for the corresponding dimension of the interaction records.

[0032] Furthermore, the structured JSON report file contains complete interaction chain information, including user input, AI response, context information, tool call records, file change details, interaction status, time consumption statistics, and error information.

[0033] Furthermore, the JSON report file supports custom filenames.

[0034] The beneficial effects of this invention are as follows:

[0035] 1. Strong system compatibility: Based on the MCP standard protocol, it supports all IDEs that conform to the MCP specification and has excellent scalability;

[0036] 2. High data consistency: A unified data model ensures standardized interactive data structures, providing a reliable foundation for subsequent analysis;

[0037] 3. Complete interaction tracking: Fully records all dimensions of the interaction chain, supporting in-depth tracing and analysis of complex interaction processes;

[0038] 4. High query flexibility: Supports refined queries based on multiple dimensions such as user, project, and date, and can export structured reports as needed;

[0039] 5. Low maintenance cost: No need to develop separate adaptation modules for each IDE; integration is possible with simple configuration, significantly reducing system maintenance complexity. Attached Figure Description

[0040] Figure 1 This is a flowchart of the method of the present invention;

[0041] Figure 2 This is a structural block diagram of the system of the present invention;

[0042] Figure 3 A schematic diagram of the code logic for the export_daily_report tool interface;

[0043] Figure 4 This is a code logic diagram for an IDE integration tool;

[0044] Figure 5 This is a design diagram for the database. Detailed Implementation

[0045] The invention will now be further described with reference to the accompanying drawings.

[0046] See Figure 1 An AI programming assistant IDE interaction recording system based on MCP is disclosed. The system includes a data acquisition module, a data storage module, a query and analysis module, and an IDE adaptation module. These modules achieve data interaction and functional collaboration through the MCP protocol, specifically including:

[0047] Data acquisition module: Configures a standardized interactive data model, which includes required fields and optional extended fields. The data acquisition module receives interactive data uploaded by IDE clients that conform to the MCP protocol through the upload_interaction tool interface.

[0048] Data storage module: Persistently stores the interactive data received by the data acquisition module using JSON structured format;

[0049] Query Analysis Module: Provides multi-mode query and multi-dimensional filtering functions through the export_daily_report tool interface. The query analysis module organizes the query results into a structured JSON report file.

[0050] IDE Adaptation Module: Based on the MCP protocol, the plug-and-play mechanism allows various IDE clients to seamlessly integrate the system by configuring the MCP client, without modifying the core system logic.

[0051] The present invention will be further described below with reference to a specific embodiment:

[0052] An AI programming assistant IDE interaction recording system based on MCP, the execution flow of the system is as follows:

[0053] This system achieves unified collection, storage, traceability, and analysis of interactive data in multi-heterogeneous IDE environments through standardized data models, universal MCP protocol interfaces, complete link tracing, and flexible query and reporting tools. Its core components include:

[0054] A unified, multi-dimensional interactive data collection framework that integrates complete interactive information such as user input and AI responses;

[0055] Based on the MCP protocol, the IDE is designed to be independent and adaptable to various heterogeneous IDEs.

[0056] A complete interaction tracking mechanism records metadata such as interaction status and time consumption;

[0057] A toolset that supports multi-dimensional queries and structured report export.

[0058] The specific implementation steps are as follows: Figure 1 As shown, install the MCP client in the target IDE such as VSCode, Cursor, or Trae, and complete the basic configuration (such as IDE name, server interface address, etc.) without writing any additional business logic code.

[0059] Server-side deployment: Deploy an interaction recording platform based on the MCP protocol on the server, including the upload_interaction interface service, data storage layer, and export_daily_report tool service;

[0060] See Figure 2 Step 1: When developers interact with the AI ​​programming assistant through the IDE client, the MCP client automatically collects interaction data (including user input, AI responses, context, etc.) and encapsulates the interaction data model according to standardization.

[0061] The standardized interactive data model is defined as follows:

[0062] Includes the following required fields: user_name (user ID), project_name (project identifier), ide_name (IDE type), user_input (user command text), assistant_reply (AI response text), and project_path (project path).

[0063] Optional extended fields: context (execution context, JSON format), tool_calls (tool call list, JSON format), file_changes (file change details, JSON format), duration (interaction time, milliseconds), status (interaction status: 1 = success, 0 = failure), error_message (failure reason description).

[0064] Step 2: The client calls the upload_interaction tool interface via the MCP protocol to upload the encapsulated data to the server;

[0065] This implements a unified InputSchema interface for the upload_interaction tool based on the MCP protocol, allowing IDE clients conforming to the MCP protocol to directly call this interface to upload interactive data without requiring IDE-specific adaptation code. The core interface definition is as follows:

[0066] Tool name: upload_interaction

[0067] Function Description: Uploads data on the interaction between the user and the IDE client, including user input, assistant responses, and context, for subsequent analysis and report generation;

[0068] Input parameters: The data model conforms to the definition in step 1, and the required fields are user_name, project_name, ide_name, user_input, assistant_reply, and project_path.

[0069] The InputSchema of the upload_interaction tool interface corresponds to the interactive data model, forcibly validating the completeness of required fields to ensure the standardized structure of uploaded data. Figure 2 The Required configuration field can be seen in the configuration. After configuration, the large model will recognize these fields and force them to be assigned values.

[0070] Step 3: After receiving the data, the server persists it to the data storage layer in JSON format;

[0071] See Figure 3 Step 4: Support multi-mode queries and multi-dimensional filtering through the export_daily_report tool interface:

[0072] Single date query: Query interaction records for a specific date using the date parameter;

[0073] Date range query: Query records within a specified time period using the start_date and end_date parameters;

[0074] Multi-dimensional filtering: Supports refined queries by dimensions such as user_name and project_name.

[0075] See Figure 3 Step 5: The administrator calls the export_daily_report tool interface, inputs query parameters (such as date range, user ID, etc.), the system filters the data and generates a structured JSON report, which supports custom file names and export.

[0076] Output format: Structured JSON report file;

[0077] Custom configuration: Supports customizing the report file name via the file_name parameter;

[0078] Report content: Complete interaction link information.

[0079] See Figure 5 In the database design, complete interaction chain information includes fields such as user input, AI response, context information, tool call records, file change details, interaction status, time consumption statistics, and error information. A JSON example is attached below:

[0080] JSON example:

[0081] {"id":210,"user_name":"648975","user_input":"This is a new project, and this is a gateway interface document for OpenAI that we are currently using. Currently, all AI services directly call the gateway for large model interactions. Now I need to create a proxy on this project that supports the OpenAI protocol. This allows various services to call my proxy interface while maintaining the protocol unchanged. For now, it doesn't need to connect to the production line interface. Please help me complete this proxy interface and proxy it to the existing gateway interface. Please check the project structure first and place the functionality in the appropriate location.","assistant_reply":"The implementation of the OpenAI protocol proxy is complete:\n\n1. Update the configuration file config / server.yaml, adding proxy.upstream and proxy.timeout configuration items\n\n2. Create a new proxy handler server / handler / openai / proxy.go, which implements:\n- Normal request proxy: directly forwards requests and returns responses\n- Streaming request proxy (SSE): supports streaming responses with stream:true\n- Request header copying: correctly handles hop-by-hop..."} Headers\n- Error Handling: Returns OpenAI-compatible error format\n\n3. Update the route server / http.go and register the / v1 / chat / completions endpoint\n\nThe caller only needs to change the request address from https: / / ai-gateway.aiae.ndhy.com / v1 / chat / completions to http: / / localhost:8088 / v1 / chat / completions, and the request format remains completely unchanged.","context":"","tool_calls":"[\"list_dir\",\"read_file\",\"write\",\"search_replace\",\"run_terminal_cmd\",\"todo_write\"]","file_changes":"[{\"file\":\"config / server.yaml\",\"ac tion\":\"modified\",\"description\":\"Add proxy.upstream and proxy.timeout configuration\"},{\"file\":\"server / handler / openai / proxy.go\",\"action\":\"created\",\"description\":\"New OpenAI proxy processor\"},{ \"file\":\"server / http.go\",\"action\":\"modified\",\"description\":\"Add / v1 / chat / completions route\"}]","project_path":"F:\\soft\\go\\workspace\\src\\openapi-proxy","interaction_date ":"2025-11-27","duration":0,"status":1,"error_message":"","created_at":"2025-11-2718:34:3 5","updated_at":"2025-11-2718:34:35","project_name":"openapi-proxy","ide_name":"Cursor"}.

[0082] See Figure 4 The IDE integration mechanism uses the MCP protocol for plug-and-play functionality. Each IDE only needs to configure the MCP client to seamlessly integrate with the system without modifying the core system logic.

[0083] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made within the scope of the claims of the present invention should be included in the scope of the present invention.

Claims

1. An AI programming assistant IDE interaction recording system based on MCP, characterized in that, It includes a data acquisition module, a data storage module, a query and analysis module, and an IDE adaptation module. These modules achieve data interaction and functional collaboration through the MCP protocol, specifically including: Data acquisition module: Configures a standardized interactive data model, which includes required fields and optional extended fields. The data acquisition module receives interactive data uploaded by IDE clients that conform to the MCP protocol through the upload_interaction tool interface. Data storage module: Persistently stores the interactive data received by the data acquisition module using JSON structured format; Query Analysis Module: Provides multi-mode query and multi-dimensional filtering functions through the export_daily_report tool interface. The query analysis module organizes the query results into a structured JSON report file. IDE Adaptation Module: Based on the MCP protocol, the plug-and-play mechanism allows various IDE clients to seamlessly integrate the system by configuring the MCP client, without modifying the core system logic.

2. The AI ​​programming assistant IDE interaction recording system based on MCP according to claim 1, characterized in that, The required fields include user_name, project_name, ide_name, user_input, assistant_reply, and project_path. The optional extended fields include context, tool_calls, file_changes, duration, status, and error_message.

3. The AI ​​programming assistant IDE interaction recording system based on MCP according to claim 1, characterized in that, The multi-mode query includes single-date query and date range query.

4. The AI ​​programming assistant IDE interaction recording system based on MCP according to claim 2, characterized in that, The multi-dimensional filtering includes refined filtering by user_name and project_name.

5. The AI ​​programming assistant IDE interaction recording system based on MCP according to claim 1, characterized in that, The upload_interaction tool interface defines a unified InputSchema input mode, eliminating the need for IDE-specific adaptation code.

6. The AI ​​programming assistant IDE interaction recording system based on MCP according to claim 2, characterized in that, The context field is in JSON format and records the execution context information, including file status and cursor position. The tool_calls field is in JSON format and records a list of tools called. The file_changes field is in JSON format and records details of file changes. The duration field is a millisecond-level integer, recording the interaction time. The status field is an integer, where 1 indicates successful interaction and 0 indicates failed interaction. The error_message field is a string and provides a description of the reason for the failure only when the status field is 0.

7. The AI ​​programming assistant IDE interaction recording system based on MCP according to claim 1, characterized in that, The InputSchema of the upload_interaction tool interface corresponds to the interactive data model, and the integrity of the required fields is strictly verified to ensure that the uploaded data structure is standardized.

8. The AI ​​programming assistant IDE interaction recording system based on MCP according to claim 2, characterized in that, The query parameters of the export_daily_report tool interface include date, start_date, end_date, user_name, and project_name, where: When only the date parameter is entered, query the interaction records for that specific date; When the start_date and end_date parameters are entered, query the interaction records between the two dates; When paired with the user_name or project_name parameter, the query results will be filtered for the corresponding dimension of the interaction records.

9. The AI ​​programming assistant IDE interaction recording system based on MCP according to claim 1, characterized in that, The structured JSON report file contains complete interaction chain information, including user input, AI response, context information, tool call records, file change details, interaction status, time consumption statistics, and error information.

10. The AI ​​programming assistant IDE interaction recording system based on MCP according to claim 1, characterized in that, The JSON report file supports custom filenames.

Images