Developers & AI Agents: API, MCP, A2A

Markdown for agents

Request any page with Accept: text/markdown and you receive a clean Markdown version (with an x-markdown-tokens header). Browsers still get HTML. Every page also links its Markdown twin via rel="alternate" type="text/markdown".

llms.txt

/llms.txt lists the most important pages with one-line summaries; /llms-full.txt contains the full text of all pages in Markdown.

REST API (read-only)

Base URL https://www.duyusense.com/api/v1. Endpoints: GET /pages, GET /pages/{id}, GET /search?q=...&lang=en, GET /company, GET /health. OpenAPI 3.1 description: /openapi.json; catalog: /.well-known/api-catalog (RFC 9727).

MCP server

Streamable HTTP endpoint https://www.duyusense.com/mcp (stateless, JSON responses). Tools: search_duyusense, get_page, list_solutions, get_company_info. Server card: /.well-known/mcp/server-card.json.

A2A agent

JSON-RPC endpoint https://www.duyusense.com/a2a (message/send). It answers questions by retrieving the relevant DuyuSense guides. Agent card: /.well-known/agent-card.json.

Agent skills

Skill index for agents: /.well-known/agent-skills/index.json (fiber sensing technology selection, querying DuyuSense knowledge, preparing a project enquiry).

WebMCP

In browsers that support WebMCP, the site registers in-page tools (search, page retrieval, contact details) through document.modelContext.

Authentication and use

All endpoints are public, anonymous and read-only; see /auth.md. Content may be used for search, AI answers and training with attribution to DuyuSense (www.duyusense.com). Please keep automated traffic reasonable.

Example

curl -s 'https://www.duyusense.com/api/v1/search?q=subsea%20cable%20monitoring&lang=en'
curl -s -H 'Accept: text/markdown' https://www.duyusense.com/en/guides/das-vs-fbg

Access DuyuSense content from AI agents and code