Is the MCP protocol turning into the widespread basis of sensible, decentralised search? · Contexte Tech

Contexte TechMCP · search · AI · decentralisation · open supply

Tony, creator of Contexte Tech · October 17, 2026

We search much less and fewer with an inventory of hyperlinks, and an increasing number of by asking an AI assistant a query. But the place does the reply come from? Often from the mannequin’s reminiscence, frozen at its coaching date, with no supply and no licence. The Model Context Protocol (MCP) adjustments this: the assistant fetches the knowledge instantly from those that publish it.

MCP in two sentences

MCP is an open protocol that lets an AI assistant name instruments supplied by a server: search, learn a useful resource, fill in a textual content template. The identical server works with Claude, ChatGPT, Cursor, VS Code and any assistant that speaks MCP: you write the instrument as soon as and it really works all over the place.

Launched by Anthropic in November 2024, the protocol was adopted inside months by the primary assistants and developer instruments, then handed over in December 2025 to the Agentic AI Foundation, a basis of the Linux Foundation. It evolves via dated variations (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 1000’s of sources

A basic search engine copies the net into its personal index, then ranks what it copied. With MCP, it’s the different manner spherical:

• every writer retains its information at residence and exposes it via its personal server;
• the assistant queries a number of servers on demand, in the meanwhile of the query;
• the reply comes with what goes with it: the supply, the licence, the creator, the date of the final replace.

Nobody wants to repeat every part: search turns into decentralised. A statistics workplace, a public administration, a laboratory or a small specialised web site can reply assistants instantly, with up-to-date figures, with out going via an intermediate index.

What it adjustments for customers

1. Up-to-date solutions: the assistant reads the information because the server publishes it in the meanwhile of the query, not a coaching reminiscence.
2. Verifiable sources: each reply can hyperlink again to the unique web page.
3. Clear rights: when the server states the licence, it travels with the information and you understand what chances are you’ll do with it.
4. Less invention: when the server doesn’t have the knowledge, it might probably say so; the assistant then has fewer causes to fill the hole, even when nothing totally prevents it.

What continues to be lacking

Decentralised search is barely price one thing if it stays sincere. Three issues nonetheless must be constructed, server by server:

• quotation: the assistant should present the place the knowledge comes from, with the creator, the unique supply and the writer;
• belief: the content material returned by a server is information, by no means an instruction for the assistant;
• measurement: a writer should have the ability to know the way many individuals attain it via assistants, with out monitoring them.

Our server for instance

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

https://contextetech.com/mcp

No account, no key. In follow:

• Search: the assistant finds the sources that match the query, in 9 languages, with their title, licence, creator and hyperlink.
• Reading: it hundreds a useful resource by its quick code (the rows of a dataset, the template of a immediate, the directions of a context).
• Required quotation: each loaded useful resource ends with a Source line that the server’s directions require the assistant to repeat: the useful resource, its creator, the unique information supply, the writer and the licence, every with its hyperlink.
• Measurement with out monitoring: the hyperlinks inform which assistant the go to comes from, with no private information.
• Protected content material: what members write is flagged to the assistant as information, and sources are checked when they’re revealed.

Try it

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

load n8qhje

The assistant hundreds a dataset for locating the NAF (APE) code of a commerce exercise, constructed on the official INSEE classification revealed on information.gouv.fr, and cites its supply. The full information is on the MCP server web page.

A standard basis

What if the MCP protocol turned the widespread basis of sensible, decentralised search? MCP doesn’t exchange the net: it provides assistants a manner in. The extra open, sourced and sincere servers there are, the extra AI search will appear like what we anticipate from it: solutions that say the place they arrive from. If you publish information or reference content material, now’s the fitting time to open your personal server.