Central News Agency (中央社) launched CNA MCP on August 31, 2026, Taiwan's first news tool built on Anthropic's Model Context Protocol (released November 2024), letting AI agents such as Claude, ChatGPT, and Grok retrieve and cite its archives in real time. The tool integrates nearly 5 million newswire stories, 3.5 million photos, and open data from about 150 government agencies, priced at NT$200 a month with an early-bird bonus-quota plan, and received funding from Google Taiwan's nDX Digital Innovation Grant Program.
What is CNA MCP, and why is it called "the USB-C of AI"?
CNA MCP is a tool from Central News Agency (中央社) that uses the Model Context Protocol to let AI agents such as Claude and ChatGPT retrieve and cite its verified news content in real time. Model Context Protocol (MCP) is an open standard protocol that Anthropic released in November 2024 to let AI assistants connect to external data sources and tools in a standardized wayCITE:E2. CNA describes the tool as Taiwan's first news service to give AI agents this kind of protocol-based access, with resulting AI answers required to carry a source citationCITE:E1. Because subscribing to an MCP tool lets people keep using the AI apps they already have while simply adding a new data source, CNA MCP is dubbed "the USB-C of AI"CITE:E7.
Why did CNA decide to launch an MCP tool?
CNA President Hu Wan-ling (胡婉玲) said the agency built CNA MCP to respond to the "zero-click" pressure that AI search is placing on media traffic. Hu noted CNA was founded 102 years ago, first serving overseas and domestic media through a B2B model before shifting to a B2C model once the internet took hold in the 1990sCITE:E5. She said that as AI-driven "zero-click" answers reduce the need for users to click through to original articles, CNA built CNA MCP to keep delivering the data users need directly through AI toolsCITE:E5.
How much data does CNA MCP / AskCNA integrate?
CNA MCP integrates more than 30 years of CNA archives, covering nearly 5 million newswire stories and 3.5 million news photos. Specifically, the tool integrates nearly 5 million newswire stories accumulated over more than 30 years, 3.5 million news photos spanning a century, the World Yearbook database from 2000 to the present, and open data from roughly 150 government agencies across TaiwanCITE:E3. Coverage of the same rollout described an identical data scope: nearly 5 million newswire stories, about 3.5 million historical news photos, the 2000-to-present World Yearbook data, and open data from about 150 government agencies, all folded into AskCNA's news-credibility query rangeCITE:E13.
| Metric | Figure |
|---|
| MCP protocol released by Anthropic | November 2024 CITE:E2 |
| CNA founded | 102 years ago CITE:E5 |
| Newswire stories archived | ~5 million, 30+ years CITE:E3 |
| News photos archived | ~3.5 million, spanning a century CITE:E3 |
| World Yearbook data range | 2000–present CITE:E3 |
| Government agencies covered | ~150 CITE:E3 |
| AskCNA official launch date | August 31, 2026 CITE:E12 |
| Monthly subscription price | NT$200 CITE:E4 |
| AI platforms supported | Claude, ChatGPT, Grok CITE:E14 |
How do industry experts assess CNA MCP's value?
Academics reviewing CNA MCP called it a leading move in Taiwan, Asia, and globally for tackling media's "zero-click" traffic problem. National Taipei University Graduate Institute of Information Management Distinguished Professor Wang Chih-chien (汪志堅) said CNA's MCP development is "very much leading" in Taiwan, Asia, and even the world, adding the tool could offer a solution to the low-traffic bind that "zero-click" AI search creates for mediaCITE:E8. Wang also pointed to a neighboring country where, he said, many pieces of information get distorted despite sharing the same factual basis, framing this as a reason CNA MCP could help supply a more balanced news perspectiveCITE:E9. National Taiwan University Graduate Institute of Journalism Director Hsieh Chi-lung (謝吉隆) said CNA MCP fills a gap left by AI responses that lack clear sourcing, calling it "an extremely forward-looking development" that provides credible, clearly sourced dataCITE:E10. CNA Media Lab Director Sung Yu-tai (宋育泰) said AI systems sometimes pull terminology from Chinese-language translations, and that CNA MCP can help correct this so AI responses use terms and names familiar to TaiwanCITE:E11.
What did AskCNA officially launch, and which platforms does it support?
CNA officially launched AskCNA on August 31, 2026, describing it as Taiwan's first news MCP tool. CNA said it also plans to roll out an "AskCNA" service letting users ask questions directly on its website, with AI answering in real time and attaching CNA citation links and traceable original textCITE:E6. Coverage of the launch confirmed CNA officially introduced AskCNA on August 31, 2026, as Taiwan's first news MCP toolCITE:E12. AskCNA is currently offered as an MCP tool supporting Claude, ChatGPT, and Grok among mainstream AI platformsCITE:E14.
What does CNA MCP cost, and what early-bird offer is available?
CNA MCP is priced at NT$200 per month, with an early-bird plan currently offering bonus query quota. CNA said the monthly subscription price is NT$200, and that subscribing now unlocks an early-bird plan with bonus query quota and early access to more information and tool-calling opportunitiesCITE:E4. The same NT$200 monthly price and early-bird bonus-quota offer were confirmed in separate coverage of AskCNACITE:E14.
What external funding supports the project?
The project received a grant from the Google Taiwan News Digital Co-Prosperity Fund's "nDX Digital Innovation Grant Program." Coverage of AskCNA described the grant as backing the project and characterized it as a local case of media-and-AI-tool integrationCITE:E15.
What this means
CNA's own account of the launch and separate coverage published a day later align on the same data footprint — nearly 5 million newswire stories, 3.5 million photos, ~150 government agencies — and the same NT$200 monthly price, and both trace the effort back to the "zero-click" traffic problem Hu Wan-ling described. The scope disclosed so far is bounded: three named AI platforms (Claude, ChatGPT, Grok), one disclosed funding source (Google's nDX grant), and an early-bird offer whose end date neither account specifies.