We just upgraded tetra.vn: the website now exposes tools to AI agents via the WebMCP standard. That means an AI agent — in the browser, or a tool like Claude Code or Cursor — can call our website to ask about products, look up knowledge, search articles or navigate, instead of just scraping HTML.
What is WebMCP?
WebMCP is an emerging web standard (W3C Web Machine Learning group) that lets a web page declare its own “tools” for AI agents to call right in the browser, via document.modelContext. It brings the spirit of the Model Context Protocol (MCP) — how AI agents call tools — directly into the page. The web is shifting: not just humans reading, but AI agents acting.
How we did it (briefly)
Very light: a small layer that feature-detects browser support (no support = skipped, affects no one), then declares a few read-only tools. No speed trade-off, no personal-data collection.
How tetra.vn differs from an ordinary website — to an AI agent
With an ordinary website, an AI agent has to scrape and guess at the HTML — slow, error-prone, easy to miss or misread. With tetra.vn, the agent calls structured tools and gets back clean, accurate data.
A concrete example: an agent asks “What products does Tetra have, and which are open source?” → it calls list_tetra_products and gets the exact list with descriptions and links — an accurate answer, nothing invented, nothing missed. Ask about the company → get_tetra_knowledge returns the full curated knowledge base.
This is the next step of AEO: we already optimized for AI to read us correctly (llms.txt, structured FAQ); now the site also lets AI call it correctly.
And it gets ahead of a wave: more and more people use AI agents to evaluate vendors. When someone tells an agent “find a company doing on-prem AI in Vietnam”, a callable website is understood and represented correctly — instead of being misread.
Fittingly, it matches Tetra's own thesis: Angler gives web access to AI agents (one end), while WebMCP makes your web callable by agents (the other end) — plus AI running on-prem for real operations.
In fairness: WebMCP is early, no browser enables it by default yet — this is a deliberate head-start, not something every agent can use today.
Want your website / systems to be “agent-ready” too?
If your business wants to be understood correctly by AI agents — from AEO to WebMCP for your website, to putting AI agents into your internal workflows (on-prem, data never leaving your systems) — Tetra can build that for you.
Book a consultation — tell us your problem and we'll propose the right approach.
Related articles

Google’s Open Knowledge Format (OKF): packaging knowledge for AI agents — what we learned deploying it
Google just published OKF — an open standard that turns organizational knowledge into a Markdown directory agents read directly. We took the samples apart and pulled out deployment lessons for several domains.
Read ↗
Angler: an open-source search, crawl & research gateway you self-host
Angler is a free, MIT open-source search + crawl + research gateway you self-host: web access for AI agents (Claude Code, Cursor…) over MCP or the Firecrawl SDK, turning the web into clean, bias-resistant markdown for research and AI agents.
Read ↗
Vietnamese OCR: turning document images into usable data (safely)
OCR only reads text; businesses need structured data. The difference between OCR and data extraction, why Vietnamese OCR is hard, the risk of free online OCR with sensitive data, and how to do it safely on your own server.
Read ↗Personal Data Protection checklist
Review your business before the law takes effect on 01/01/2026.
