Skip to content
中

Xiangyan AI explores applied agents and turns ideas into systems that run in real workflows—research, product engineering, and deployment in one path.

Get in touch

Product 01In build

Xiangyan Agent Workspace

A multi-agent workspace for real-world use: model access, collaborative chat, retrieval, tool execution, and admin—deployable and extensible.

Built for a stable chat experience, a unified model entry point, and lightweight infrastructure that plugs into your stack.

Get in touch
  • Agent · RAG · Tools
  • API · Webhook · Workflow
  • Trace · Eval · Audit
  • Deploy · Integrate · Operate

Capabilities

From orchestration to ops

Agent

Agent orchestration

Multi-step reasoning, tool calling, human-in-the-loop, reviewable traces.

RAG

RAG retrieval

Chunking, embeddings, semantic search, evidence records, live knowledge.

Integrate

System integration

APIs, webhooks, workflow engines—verify, fill, notify, write back.

Modal

Multimodal processing

Text, voice, documents, and images for support and internal collab.

Eval

Eval & observability

Tracing, quality evals, usage ledgers, structured logs, and alerts.

Ops

Security & ops

Access boundaries, redaction, audit logs, 2FA/SSO-ready governance.

About

Research and engineering, in public

We study new models and orchestration methods, and turn experiments into projects that run, get used, and keep evolving across research and applied practice.

  • 01Ship-ready

    Start from real workflows. Deliver agents that run—not slideware.

  • 02Integrable

    CRM, tickets, knowledge bases, internal APIs—inside your stack.

  • 03Evolvable

    Eval, observability, and feedback so systems grow with the business.

Approach

Three steps to operate

  1. 01

    Discover

    High-value flows, data boundaries, success metrics—first shippable case.

  2. 02

    Build

    Orchestrate agents, wire knowledge and systems, iterate on eval sets.

  3. 03

    Operate

    Gradual rollout, monitoring, alerts—agents in daily operations.

Scenarios

Where agents land

Internal

Knowledge assistant

Policies, projects, tickets—retrieve and draft in natural language.

CX

Support agent

Multi-turn dialog, ticket creation, escalation—consistent and auditable.

Ops

Process automation

Cross-system orchestration: verify, fill, notify, write back.