Shanghai Wangshan Releases Hitpoint XAgent to Build Full-Chain Closed Loop for Enterprise ERP

October 20 news: Recently, Shanghai Wangshan Information Technology Co., Ltd. (Shanghai Wangshan) officially released its enterprise-grade ERP Agent product, Hitpoint XAgent. The product is designed for enterprises deploying mainstream ERP systems such as Oracle NetSuite, focusing on scenarios including business report generation, business data verification, batch processing, and custom development. It is an AI execution system specifically built to deliver end-to-end task results within the scope of enterprise authorization. According to the company, unlike general-purpose AI Agents on the market that require employees to install locally on work devices, Hitpoint XAgent adopts a centralized enterprise-side deployment: it only needs to be deployed on the enterprise server, and employees can access it via a web page without installing a client on individual office devices. All operational permissions, skill configurations, knowledge bases, and runtime rules can be managed uniformly by the enterprise, and the permissions available to the AI are consistent with the employee's original system permissions. It is reported that Hitpoint XAgent features three major distinctive capabilities: First, it has an ERP industry knowledge base built by the Hitpoint Cloud team based on ERP implementation experience with 400 clients, combined with a customized knowledge base for the customer's ERP system and supported by top-tier large model capabilities, which greatly avoids AI hallucinations and makes AI output results authentic and reliable. Second, critical operations support manual approval, with full-chain invocation traces; the permissions available to the AI are aligned with the employee's original system permissions and can be configured and revoked by role. Third, skills and business knowledge accumulated by employees are uniformly retained in the enterprise knowledge base and do not leave with personnel turnover. In terms of product capabilities, Hitpoint XAgent has built an execution closed loop around actual enterprise ledger scenarios—'query, do, verify, record, run'—rather than remaining at the conversational level. Users submit analysis requests in natural language in the dialog box, and XAgent supports automatically invoking query tools such as SuiteQL, automatically completing data cleaning, multi-table joins, and aggregate calculations, generating visual results such as pie charts, bar charts, trend charts, and pivot tables, and can export XLSX and CSV files containing multiple sheets with one click, replacing the repetitive work of the traditional 'export Excel—manual table assembly—re-enter into the system.' Once an analysis confirms the definitions and reports, they can be saved as SavedQuery reports and refreshed with one click via natural language next time. After task execution, the system reads real business data again to verify the results rather than relying on the model's self-report. When critical operations such as write, delete, and deployment are involved, the system first explains the execution object, action, scope, and purpose, and continues only after manual one-click online confirmation. Every tool invocation is fully traced; the permissions available to the AI are strictly aligned with the employee's original system permissions and can be configured by role and revoked at any time, meeting enterprise internal control and audit requirements. For reusable rules, field definitions, and business understanding formed during tasks, the system proactively asks whether to write them into the enterprise's long-term memory; even after an employee leaves, these knowledge assets remain on the enterprise side, helping the enterprise continuously improve its business capabilities. Official information shows that Hitpoint XAgent now supports integration with mainstream large models such as OpenAI GPT, Claude, Gemini, DeepSeek, Kimi, GLM, and Qwen, and customer privately deployed models can also be integrated. This article was first published on the official website of Ebrun. [Source: Ebrun Go. Ebrun's automated writing robot, delivering e-commerce intelligence via algorithms in real time. This dog is still young; welcome to contact run@ebrun.com or leave a message to help it grow.]