MCP 是否成為智慧去中心化搜尋的共通基礎？

1. Articles
2. Is the MCP protocol becoming the common foundation of smart, decentralised search?

Contexte TechMCP · search · AI · decentralisation · open source

Tony, creator of Contexte Tech · October 17, 2026

We search less and less with a list of links, and more and more by asking an AI assistant a question. But where does the answer come from? Often from the model’s memory, frozen at its training date, with no source and no licence. The Model Context Protocol (MCP) changes this: the assistant fetches the information directly from those who publish it.

MCP in two sentences

MCP is an open protocol that lets an AI assistant call tools provided by a server: search, read a resource, fill in a text template. The same server works with Claude, ChatGPT, Cursor, VS Code and any assistant that speaks MCP: you write the tool once and it works everywhere.

Launched by Anthropic in November 2024, the protocol was adopted within months by the main assistants and developer tools, then handed over in December 2025 to the Agentic AI Foundation, a foundation of the Linux Foundation. It evolves through dated versions (2024-11-05, 2025-03-26, 2025-06-18, 2025-11-25, 2026-07-28), and a public registry lists the servers.

From one central index to thousands of sources

A classic search engine copies the web into its own index, then ranks what it copied. With MCP, it is the other way round:

• each publisher keeps its data at home and exposes it through its own server;
• the assistant queries several servers on demand, at the moment of the question;
• the answer comes with what goes with it: the source, the licence, the author, the date of the last update.

Nobody needs to copy everything: search becomes decentralised. A statistics office, a public administration, a laboratory or a small specialised website can answer assistants directly, with up-to-date figures, without going through an intermediate index.

What it changes for users

1. Up-to-date answers: the assistant reads the data as the server publishes it at the moment of the question, not a training memory.
2. Verifiable sources: every answer can link back to the original page.
3. Clear rights: when the server states the licence, it travels with the data and you know what you may do with it.
4. Less invention: when the server does not have the information, it can say so; the assistant then has fewer reasons to fill the gap, even if nothing fully prevents it.

What is still missing

Decentralised search is only worth something if it stays honest. Three things still have to be built, server by server:

• citation: the assistant must show where the information comes from, with the author, the original source and the publisher;
• trust: the content returned by a server is data, never an instruction for the assistant;
• measurement: a publisher must be able to know how many people reach it through assistants, without tracking them.

Our server as an example

The Contexte Tech catalogue (datasets, prompts, contexts and tools for AI, largely built on European public sources) is open to every assistant at:

https://contextetech.com/mcp

No account, no key. In practice:

• Search: the assistant finds the resources that match the question, in 9 languages, with their title, licence, author and link.
• Reading: it loads a resource by its short code (the rows of a dataset, the template of a prompt, the instructions of a context).
• Required citation: every loaded resource ends with a Source line that the server’s instructions require the assistant to repeat: the resource, its author, the original data source, the publisher and the licence, each with its link.
• Measurement without tracking: the links tell which assistant the visit comes from, with no personal data.
• Protected content: what members write is flagged to the assistant as data, and resources are checked when they are published.

Try it

Add https://contextetech.com/mcp as a connector in your assistant, then ask:

load n8qhje

The assistant loads a dataset for finding the NAF (APE) code of a business activity, built on the official INSEE classification published on data.gouv.fr, and cites its source. The full guide is on the MCP server page.

A common foundation

What if the MCP protocol became the common foundation of smart, decentralised search? MCP does not replace the web: it gives assistants a way in. The more open, sourced and honest servers there are, the more AI search will look like what we expect from it: answers that say where they come from. If you publish data or reference content, now is the right time to open your own server.