An intelligent agent creation platform for universities, enabling every teacher and student to build their own agent
From agent creation and prompt configuration to multimodal knowledge bases with RAG retrieval, visual workflow orchestration, conversation debugging and evaluation, and permission-based approval and publishing — the platform covers the full lifecycle of an agent. Teachers and students can get started without any programming background, and their work can be published to the campus portal, APIs and mobile devices. Fully private, on-premises deployment keeps data on campus. Mainstream large models such as ERNIE Bot, Tongyi Qianwen, iFlytek Spark, DeepSeek and Zhipu, as well as local campus models, are connected through a unified OpenAI-compatible protocol and can be switched with one click. The platform integrates with the university's unified identity authentication, data center, messaging and portal.
Why let teachers and students build their own agents?
Turn high-frequency, repetitive tasks such as course Q&A and policy lookup into reusable agents built once and shared across the university.
Structure subject knowledge and business experience into agents instead of leaving them as personal experience.
Only by building one yourself can you understand the model's limits and where human oversight is essential.
Eight core capabilities covering the full agent lifecycle
From creation to operation, every step is supported by a corresponding platform capability. Teachers and students focus only on their scenarios — the platform handles the rest.
Agent Building
Supports conversational, workflow-orchestration and other agent types, with sectioned configuration and real-time preview debugging.
Prompt Factory
Template-based programming, variable reference syntax and a built-in template library, with AI prompt optimization.
Long-term Memory
Contextual memory, intent recognition, conversation interruption and state recovery.
Skills & Plugin Center
Extend capabilities with plugins and MCP; skills can be assembled into different agents on demand.
From type selection to prompt optimization — get your agent right the first time
The builder is organized into seven configuration sections — model selection, prompt writing, memory, skills, knowledge, conversation and content safety review — with a built-in debug preview pane on the right so you can test as you edit.
Supports conversational, workflow-orchestration and other agent types — just pick the one that fits your scenario.
Customize or link a knowledge base, with preset opening lines and guide questions; after answering, multiple follow-up suggestions are pushed automatically based on context, with fallback replies configurable.
Template-based prompt programming: embed form variables and contextual information through variable reference syntax, with a built-in prompt template library ready to use.
AI prompt optimization: automatically turns natural-language descriptions into clearly structured, well-constrained prompt templates, greatly lowering the barrier for teachers and students.
Connect with existing campus systems, not another data silo
Unified Identity Authentication
Complete SSO single sign-on according to the university's interface specifications; teachers and students enter from the portal in one step and gain full system permissions without re-registering or logging in again.
Data Integration
Complete full data synchronization to the university data center using the methods (API / table / intermediate database) and formats (CSV / DAT / DBF / MDB / ODB++, etc.) specified by the university; open all database permissions and deliver a complete data dictionary, syncing any data already in the data center from it.
Unified Messaging Integration
Notifications and messages sync to the portal in real time; clicking them in the portal takes users straight to the corresponding page.
Fragmented Feature Integration
Integrated into the portal's PC and mobile ends as cards and icon links, precisely jumping to the corresponding agent or service entry.
Embedded Mobile Access
Accessed embedded within the campus mobile app, so teachers and students need no additional app and use the same account system as the portal.
Performance metrics and deployment options, stated openly
≥ 90%: Knowledge Q&A accuracy, traceability accuracy and document summarization similarity all no lower than 90%
≤ 10s: Document parsing response for 1MB no higher than 10 seconds
Data stays on campus: Fully private on-premises deployment; business data flows within the campus environment
Single-node / cluster: Adapts to the university's CPU / GPU servers, with deployment form chosen by scale