A method for constructing an intelligent RPA platform based on a model context protocol

Abstract

The application discloses a kind of intelligent RPA platform construction methods based on model context protocol, the construction method includes: user triggers;Agent planning and calling;MCP routing and execution;Result returns and context updates;AIAgent feeds back the result to user in natural language form.The application provides a new type of intelligent automation platform architecture with model context protocol (MCP) as the core, which can deeply integrate large model cognitive ability and RPA execution ability.

Description

Technical Field

[0001] This invention relates to the field of robotic process automation, and more particularly to a method for constructing an intelligent RPA platform based on a model context protocol. Background Technology

[0002] Robotic Process Automation (RPA) technology uses software robots to simulate human operations on a user interface and has been widely used to process large amounts of repetitive, well-defined structured data tasks, achieving cost reduction and efficiency improvement in business processes. However, traditional RPA has obvious limitations: first, the processes are highly rigid, and once the source or target system changes, the automated processes are very prone to failure; second, it lacks cognitive and decision-making capabilities and cannot handle unstructured data or scenarios requiring complex judgments.

[0003] In recent years, the rise of Large Language Models (LLMs) has brought new opportunities to solve the aforementioned problems. Large models possess powerful semantic understanding, reasoning, and generation capabilities, but they lack standardized connection channels between themselves and complex internal business systems, data sources, and tools. To bridge this gap, the Model Context Protocol (MCP) has emerged. MCP is an open standard protocol proposed and open-sourced by Anthropic, aiming to establish standardized connections between large models and the external world (such as data, tools, and APIs), acting similarly to a "USB-C interface" for AI applications. MCP adopts a client-server architecture, enabling large models to call external functions or request structured data in a consistent manner by defining a unified communication protocol (usually based on JSON-RPC).

[0004] Currently, the closest existing technical solutions mainly fall into two categories: Option 1: Traditional RPA platform. Its typical architecture consists of three core parts: the development and design layer (for process orchestration), the control and management layer (for scheduling and monitoring), and the robot execution layer (for simulating UI operations). This option relies entirely on preset rules and has a low level of intelligence.

[0005] Option 2: Initial Integration of Large Models and RPA. Some vendors have begun experimenting with using large models as the "brain," calling RPA robots to perform specific operations through proprietary or non-standard API interfaces. For example, Daguan RPA has announced compatibility with the MCP protocol and integration with Baidu Smart Cloud Qianfan platform, allowing RPA processes to be called through MCPServer. However, these solutions are mostly point-to-point integrations, lacking a systematic platform-level reconstruction centered on MCP, and failing to fully leverage the potential of MCP as a standardized capability layer.

[0006] Disadvantages of existing technology: Traditional RPA lacks intelligence: it can only handle rule-based, deterministic tasks and cannot cope with ambiguous, unstructured inputs or scenarios that require dynamic decision-making in business processes.

[0007] Non-standard integration methods: Existing attempts to combine large models with RPA often use customized, non-standard interfaces, resulting in high system coupling, poor maintainability and scalability.

[0008] Weak context management capabilities: Traditional solutions lack effective management and transmission mechanisms for dynamic context during task execution, which limits the realization of complex, multi-step automated tasks. Summary of the Invention

[0009] In view of the above problems, the present invention is proposed to provide a method for building an intelligent RPA platform based on a model context protocol to overcome or at least partially solve the above problems.

[0010] According to one aspect of the present invention, a method for building an intelligent RPA platform based on a model context protocol is provided, the method comprising: User-triggered; Agent planning and invocation; MCP routing and execution; Results returned and context updated; AIAgent will provide the results to the user in natural language.

[0011] Optionally, the user trigger specifically includes: the user submitting a natural language task request to AIAgent.

[0012] Optionally, the Agent planning and invocation specifically includes: After understanding the task, AIAgent plans the execution steps based on its internal knowledge and the list of available MCP tools; The MCPClient initiates a series of tool call requests to the MCPBroker.

[0013] Optionally, the MCP routing and execution specifically includes: For operations that require RPA execution, the request is routed to RPACoreServer; After receiving an MCP request, RPACoreServer parses it into internal, executable automated process instructions and schedules an idle RPARobot to execute it. For other operations, the request may be routed to the database query tool server.

[0014] Optionally, the operations that need to be performed by RPA specifically include: opening an email client, reading attachments, and entering data into the ERP system.

[0015] Optionally, the result return and context update specifically include: After each MCPServer completes its task, it returns the structured results to AIAgent via the MCP protocol. The agent updates its internal context based on the returned results and decides on the next operation until the entire task is completed.

[0016] Optionally, the MCPBroker, as the central hub of the MCP ecosystem, receives requests from AIAgent and routes them to the corresponding MCPServer according to the request content; it is responsible for handling general functions such as authentication, authorization, and load balancing.

[0017] Optionally, the MCPServer is a standardized encapsulation of various capabilities; The core functions of traditional RPA platforms, including process orchestration and robot scheduling, are encapsulated into one or more RPACoreServers and registered as MCPServers. Other commonly used tools within the enterprise, including databases and email systems, are also packaged as independent MCPServers.

[0018] This invention provides a method for constructing an intelligent RPA platform based on the Model Context Protocol (MCP). The method includes: user triggering; agent planning and invocation; MCP routing and execution; result return and context update; and the AIAgent feeding back the results to the user in natural language. It provides a novel intelligent automation platform architecture with the MCP as its core, capable of deeply integrating the cognitive capabilities of large models with RPA execution capabilities.

[0019] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1A flowchart illustrating a method for constructing an intelligent RPA platform based on a model context protocol, provided in an embodiment of the present invention. Detailed Implementation

[0022] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0023] The terms "comprising" and "having," and any variations thereof, in the specification, embodiments, claims, and drawings of this invention are intended to cover non-exclusive inclusion, such as including a series of steps or units.

[0024] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0025] This invention aims to provide an intelligent RPA platform and its construction method based on the Model Context Protocol (MCP). By using MCP as the core communication and integration standard, the traditional RPA platform is reconstructed to achieve the following objectives: It endows the RPA platform with powerful cognitive and decision-making capabilities, enabling it to handle unstructured data and complex business logic.

[0026] Establish a standardized, loosely coupled integration framework to simplify the connection between large models and RPA robots and other enterprise tools.

[0027] Enables effective management and transmission of dynamic context, supporting more complex and intelligent end-to-end automated processes.

[0028] like Figure 1 As shown, the core idea of ​​this invention is to use the MCP protocol as the central nervous system of the entire platform and build a layered, modular intelligent RPA platform around it.

[0029] The platform mainly includes the following components: Platform Overall Architecture Diagram and Description Architecture Description: AIAgent (Large Model): Serves as the intelligent hub of the entire automation process. It sends tool invocation requests to the MCP ecosystem via the built-in MCPClient.

[0030] MCPBroker: As the central hub of the MCP ecosystem, it receives requests from AIAgent and routes them to the corresponding MCPServer based on the request content. It is also responsible for handling general functions such as authentication, authorization, and load balancing.

[0031] MCPServers are standardized encapsulations of various capabilities. This invention encapsulates the core functions of traditional RPA platforms (such as process orchestration and robot scheduling) into one or more RPACoreServers and registers them as MCPServers. Simultaneously, other commonly used enterprise tools (such as databases and email systems) can also be encapsulated as independent MCPServers.

[0032] RPARobots: These are the final execution units, receiving instructions from the RPACoreServer to complete specific automated operations.

[0033] Core Workflow The workflow of this invention is as follows: User trigger: The user submits a natural language task request to AIAgent (e.g., "Please process all the invoices I received today and enter the results into the ERP system").

[0034] Agent Planning and Invocation: After understanding the task, AIAgent plans the execution steps based on its internal knowledge and the list of available MCP tools. It then initiates a series of tool invocation requests to MCPPBroker through MCPClient.

[0035] MCP routing and execution: For operations that require RPA to perform (such as "open email client", "read attachment", "enter ERP data"), the request is routed to RPACoreServer.

[0036] After receiving an MCP request, RPACoreServer parses it into internal, executable automated process instructions and schedules an idle RPARobot to execute them.

[0037] For other operations (such as "query supplier information"), the request may be routed to the database query tool server.

[0038] Result Return and Context Update: After each MCPServer completes its task, it returns the structured results to the AIAgent via the MCP protocol. The Agent updates its internal context based on the returned results and determines the next operation until the entire task is completed.

[0039] Final feedback: AIAgent will provide the final results to the user in natural language.

[0040] Key technological innovations MCP-based encapsulation of RPA capabilities: Abstracting and encapsulating atomic operations (such as clicks, inputs, and screen readings) and complex processes (such as "processing invoices") of traditional RPA platforms into tools that conform to the MCP standard, enabling them to be directly discovered and invoked by large models.

[0041] Dynamic context-driven process orchestration: Traditional RPA is driven by static flowcharts, while this platform is driven by dynamic context. AIAgent dynamically determines which MCP tools to invoke based on the task objectives and the context information returned in real time, achieving true intelligent decision-making and process adaptation.

[0042] Beneficial effects: It overcomes the shortcomings of traditional RPA platforms, such as low intelligence, non-standardized integration methods and weak context management capabilities, and provides a new intelligent automation platform architecture with Model Context Protocol (MCP) as the core, which can deeply integrate the cognitive ability of large models and the execution capability of RPA.

[0043] The above specific embodiments further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for constructing an intelligent RPA platform based on a model context protocol, characterized in that, The construction method includes: User-triggered; Agent planning and invocation; MCP routing and execution; Results returned and context updated; AIAgent will provide the results to the user in natural language.

2. The method for constructing an intelligent RPA platform based on a model context protocol according to claim 1, characterized in that, The user trigger specifically includes: the user submitting a natural language task request to AIAgent.

3. The method for constructing an intelligent RPA platform based on a model context protocol according to claim 1, characterized in that, The Agent planning and invocation specifically includes: After understanding the task, AIAgent plans the execution steps based on its internal knowledge and the list of available MCP tools; The MCPClient initiates a series of tool call requests to the MCPBroker.

4. The method for constructing an intelligent RPA platform based on a model context protocol according to claim 1, characterized in that, The MCP routing and execution specifically include: For operations that require RPA execution, the request is routed to RPACoreServer; After receiving an MCP request, RPACoreServer parses it into internal, executable automated process instructions and schedules an idle RPARobot to execute it. For other operations, the request may be routed to the database query tool server.

5. The method for constructing an intelligent RPA platform based on a model context protocol according to claim 4, characterized in that, The specific operations that require RPA to perform include: opening an email client, reading attachments, and entering data into the ERP system.

6. The method for constructing an intelligent RPA platform based on a model context protocol according to claim 1, characterized in that, The result return and context update specifically include: After each MCPServer completes its task, it returns the structured results to AIAgent via the MCP protocol. The agent updates its internal context based on the returned results and decides on the next operation until the entire task is completed.

7. The method for constructing an intelligent RPA platform based on a model context protocol according to claim 3, characterized in that, As the central hub of the MCP ecosystem, MCPBroker receives requests from AIAgent and routes them to the corresponding MCPServer based on the request content; it is responsible for handling general functions such as authentication, authorization, and load balancing.

8. The method for constructing an intelligent RPA platform based on a model context protocol according to claim 6, characterized in that, The MCPServer is a standardized encapsulation of various capabilities; The core functions of traditional RPA platforms, including process orchestration and robot scheduling, are encapsulated into one or more RPACoreServers and registered as MCPServers. Other commonly used tools within the enterprise, including databases and email systems, are also packaged as independent MCPServers.

Images