A closed loop from build to evaluation, plus two products that reach the desk and the control room
PanXing provides the runtime and governance base; PanLing and PanZhi serve business users and knowledge administrators; PanJian governs trustworthy admission. Stone brings agents to the desk, and ZhiNet brings reasoning into the monitoring network. All products share one identity and runtime system and can be bought separately or as a suite.
PanXing · Agent Platform & AI Control Center
Every business scenario shares one runtime and common capabilities, so models, memory and permissions are not integrated again for each project. As the enterprise AI control plane, models, data and systems are connected through one entry point, with tokens and budgets attributed by person, team and project — every call auditable, every result traceable.
- For
- Platform administrators, security and operations
- Form
- On-premises platform + control centre
PanLing · Conversational Portal
One conversational entry point for business users: no system training required — describe the need in natural language and the query, analysis or report is produced. The reasoning trace is visible and reviewable step by step, and results export in one click. Agent capability is operated as an app shelf, ready to use.
- For
- Business staff, frontline crews
- Form
- Conversational workbench + app marketplace
PanZhi · Knowledge Workbench
Scattered enterprise documents become reusable knowledge assets: 15 formats ingested with one click, structured chunking that preserves heading hierarchy, and automatic semantic indexing and term extraction. Each answer carries a paragraph-level citation chain back to the source, and permissions follow the document.
- For
- Knowledge administrators, domain experts
- Form
- Knowledge workbench, on-premises capable
PanJian · Trust Evaluation
The admission gate before launch and the risk check during operation — every production agent works with a licence. 20-dimension evaluation and risk grading mean no release below standard; prompt injection, privilege escalation and anomalous operations are detected in real time, with high-risk behaviour blocked and logged.
- For
- Security, compliance and platform administrators
- Form
- Evaluation service + evaluation platform
Two products that land directly in the scenario
One sits on the desk and operates software for people; the other watches the monitoring network and triages alerts for people. Both reuse platform capability, delivered in different form factors.
Stone · Embodied Office Assistant
More than Q&A: it reads the screen, operates software and brings the knowledge base and agents to the desk. Capabilities are configured on the agent platform, so changing scenario does not mean changing hardware.
- For
- Managers, project owners and functional roles
- Form
- Desktop robot (prototype in tooling)
ZhiNet · Alert Governance Platform
For monitoring networks: clustering removes duplicate alerts, rules and large models triage each one, and every action is logged. During the pilot, weekly alerts fell by 77.2%.
- For
- Security operations, on-duty analysts and auditors
- Form
- On-premises deployment (pilot stage)
How we deliver
The same product set is combined according to the customer's deployment and compliance requirements, predominantly on-premises.
Put your business in the hands of a trustworthy agent
Book a demo or email us directly — we prepare materials around your industry scenario.