legacy-use: adding AI automation interfaces to legacy software without APIs

legacy-use is an open source tool whose core role is to provide a modern REST API interface to old desktop software that has no API interface (often called "legacy software"). It uses an AI intelligence to "observe" the software's graphical user interface (GUI) and simulate a human user's keystrokes. This allows developers to send simple commands to automate old, closed software just as they would a normal API. This tool solves the pain point of a large number of organizations that still rely on critical but outdated software for their core business, but are unable to automate it. It provides a bridge for these legacy systems to access modern automated workflow and AI technologies.

Function List

• Intelligent Body API Layer: Provide an AI-driven API layer for any desktop application (Windows, Linux, etc.) to be called programmatically.
• Simulation of human operation: Recognize interface elements through AI visual models, simulate mouse clicks and keyboard inputs, and complete various complex operations.
• Connectivity and Implementation: Built-in ability to handle connections to remote desktop protocols such as RDP, VNC, etc., and manage task queues to execute commands sent to the target system.
• data extraction: Intelligentsia can recognize and grab the required data from a software interface, such as extracting numbers from a report, or copying customer information.
• REST API encapsulation: All automated operations and data extraction functions are unified and encapsulated into a standard REST API interface for easy integration with other systems.
• Monitoring and manual interventionProvides full logging and monitoring functionality with "guardrails" that allow automated processes to pause and notify a human operator to intervene in the event of an error or unforeseen circumstance.

Using Help

legacy-use is designed to provide an automated solution for legacy desktop software that does not provide an API (Application Programming Interface). It works by deploying an AI intelligence that can "see" and "operate" the software's graphical interface like a human. Here are the details of how it works and how to use it:

Architecture Core

The legacy-use architecture builds on Anthropic's "Computer Use" model and extends its capabilities to legacy software on Windows, Linux, and other operating systems. Instead of directly hacking or modifying the code of legacy software, it creates an interaction layer outside of it, consisting of the following components:

1. Connection Infrastructure
   • remote connection: The system first needs to connect to the computer running the target software. legacy-use supports standard remote connection protocols such asRDP (Remote Desktop Protocol) and VNC (Virtual Network Computing), and can be configured to use a network via theVPN (Virtual Private Network) for a secure connection.
   • task queue: All operation commands sent to the target software enter a queue system. This queue is responsible for managing the execution order and status of tasks to ensure orderly and reliable operations.
2. AI Agent Execution
   • receive instructions: You can tell the AI intelligence what tasks it needs to accomplish by sending advanced commands (Prompt). For example, "Open the 'Customer Management' program, search for 'Zhang San' and copy his phone number."
   • Visual analysis and manipulation: The intelligence receives commands and analyzes the screen image of the target computer in real time. It uses computer vision techniques to recognize interface elements such as windows, buttons, input boxes, and text.
   • Analog Keyboard and Mouse: Based on the recognition results and command requirements, the intelligent body will simulate a series of mouse movements, clicks and keyboard input operations. For example, it will move the mouse cursor to the "Search" button and click on it, then type "Zhangsan" in the input box. This process is more fault-tolerant than traditional UI automation tools (such as UIPath) because it does not rely on a fixed interface layout, but rather on visual understanding to locate elements.
3. Data Extraction & API Exposure
   • data crawl: When it is necessary to obtain information from the software, the intelligent body can extract data from a specified area on the screen or from a text field in accordance with an instruction.
   • Conversion to REST API: The most critical step is that legacy-use encapsulates the entire flow of operations, including data input and output, into a simple REST API interface. This means that your other applications can now trigger a complex set of operations from the old software with a single HTTP request and receive the returned data. For example, you can create an API endpointPOST /api/getCustomerPhoneWhen this interface is called, it automatically completes the entire process of logging in, searching, and copying numbers as described above, and returns the phone number.
4. Monitoring & Guardrails
   • Logging: Every step performed by the intelligences is recorded in detail, including screen shots and operation logs. This provides a basis for debugging and auditing.
   • Artificial intervention mechanismsAutomated processes are not 100% perfect. The system triggers a "guardrail" mechanism when the intelligence encounters an unrecognizable interface, a program crash, or an unexpected error. At this point, the process is paused and an alert is sent to a designated human operator, who can take over and manually complete the current step before handing control back to the intelligence.

Example of operation flow

Suppose you need to automatically generate compliance reports from a 2009 accounting software:

1. Configuring Connections: First, configure legacy-use to connect to the Windows server running the accounting software'sRDPAddress and credentials.
2. Defining Task Instructions: Write a clear instruction (Prompt) describing the complete steps for generating a report: "Log in to the system, click on 'Report Center', select 'Compliance Reports', set the date range to this month, click on the 'Generate' button, wait for the report to appear, click 'Export to PDF' and save the file to the 'Reports' folder on your desktop. "
3. Creating API Endpoints: Define this task as an API endpoint in legacy-use, for example /api/generateComplianceReportThe
4. invoke an API: Now, your main business system can call this API regularly at the beginning of each month. legacy-use's AI intelligences will automatically log into the server, open the accounting software, and step through all the clicks and entries like a skilled employee, eventually saving the report to a designated location.

In this way, tasks that would have required half an hour of manual labor are now automated in minutes with a single API call.

application scenario

1. Finance and Accounting Automation
   An accounting firm needed to manually generate compliance reports for each client from a desktop financial software released in 2009. This process was cumbersome and error prone. Using legacy-use, they encapsulated the entire report generation process into an API, which their new system now automatically calls to generate reports for all clients in batch, with no manual intervention required.
2. Healthcare management system process optimization
   A healthcare provider organization relies heavily on an aging Electronic Health Record (EHR) system for day-to-day management (e.g. patient information entry, billing processing). The system had no API, making it difficult to synchronize information between different departments. They deployed legacy-use to automate 30%'s administrative tasks. For example, when the new system receives a patient's appointment, it automatically calls legacy-use to create the corresponding record in the EHR system, reducing the data entry workload of the front desk staff.
3. Manufacturing inventory management
   The factory's inventory management software runs on a local Windows XP machine and requires employees to manually check inventory levels and update them to a spreadsheet on a daily basis. With legacy-use, an API can be created that allows an intelligent body to open the inventory software at regular intervals each day, query the quantities of key materials, and automatically write the results to an online database or shared table for real-time monitoring of inventory levels.
4. Bridging traditional software and modern cloud services
   A company wanted to take data from its old customer relationship management (CRM) desktop software and synchronize it to the Salesforce platform in the cloud. With legacy-use, an intelligence can be set up to periodically query the old CRM for new customer data and write that data to the cloud via Salesforce's API, thus bridging the data silos between the old and new systems.

QA

1. What is the difference between legacy-use and RPA tools like UIPath?
   The main difference between legacy-use and traditional RPA tools (e.g., UIPath) is the way it is driven by its core. Traditional RPA usually relies on predefined rules and fixed positions of interface elements (e.g. XPath), which can cause the automation process to fail if the interface changes slightly. On the other hand, legacy-use is based on AI visual modeling, which "understands" the content on the screen like a human being, and is more adaptable to changes in the interface layout, and therefore more fault-tolerant.
2. Is this tool open source? Where can I find it?
   Yes, legacy-use is an open source project. You can find its source code repository on GitHub at https://github.com/legacy-use/legacy-useThe
3. What kind of technical background is required to use this tool?
   Using legacy-use requires some technical background. You need to understand how to deploy and configure a service, understand the basic concepts of REST APIs, and be able to write clear instructions (Prompts) to direct AI intelligences. It is primarily aimed at developers or technical teams looking to integrate legacy systems into modern workflows.
4. Which software on which operating systems does legacy-use support?
   According to the project description, it is designed as a general purpose solution that can add an API layer to any desktop application, including software on Windows, Linux, and other operating systems. It connects via the Remote Desktop Protocol (RDP/VNC), so theoretically any desktop environment that can be accessed remotely can be supported.