- Articles
- Is the MCP protocol becoming the average basis of smart, decentralised search?
Contexte TechMCP · hunt · AI · decentralisation · open source
Tony, creator of Contexte Tech · October 17, 2026
We hunt small and small alongside a catalog of links, and additional and additional by asking an AI aide a question. But anywhere does the answer arrive from? Often from the model’s memory, icy at its training date, alongside no origin and no licence. The Model Context Protocol (MCP) changes this: the aide fetches the data immediately from those who publish it.
MCP in two sentences
MCP is an open protocol that lets an AI aide call tools provided by a server: search, peruse a resource, inhabit in a content template. The identical server plant alongside Claude, ChatGPT, Cursor, VS Code and any aide that speaks MCP: you compose the tool formerly and it plant everywhere.
Launched by Anthropic in November 2024, the protocol was adopted inside months by the chief assistants and developer tools, afterward handed complete in December 2025 to the Agentic AI Foundation, a basis 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 community register lists the servers.
From one chief indicator to thousands of sources
A traditional hunt motor copies the web into its own index, afterward ranks what it copied. With MCP, it is the another way round:
- each publishing company keeps its data at residence and exposes it through its own server;
- the aide queries multiple servers on demand, at the instant of the question;
- the answer comes alongside what goes alongside it: the source, the licence, the author, the date of the final update.
Nobody needs to copy everything: hunt becomes decentralised. A data office, a community administration, a lab or a small specialised website can answer assistants directly, alongside up-to-date figures, without going through an intermediate index.
What it changes for users
- Up-to-date answers: the aide says the data as the server publishes it at the instant of the question, not a training memory.
- Verifiable sources: all answer can nexus rear to the first page.
- Clear rights: whenever the server states the licence, it travels alongside the data and you cognize what you may do alongside it.
- Less invention: whenever the server does not have the information, it can say so; the aide afterward has small reasons to inhabit the gap, equal if nothing completely prevents it.
What is motionless missing
Decentralised hunt is lone value item if it stays honest. Three things motionless have to be built, server by server:
- citation: the aide must display anywhere the data comes from, alongside the author, the first origin and the publisher;
- trust: the satisfied returned by a server is data, never an education for the assistant;
- measurement: a publishing company must be capable to cognize how many group attain it through assistants, without tracking them.
Our server as an example
The Contexte Tech catalogue (datasets, prompts, contexts and tools for AI, mostly built on European community sources) is open to all aide at:
https://contextetech.com/mcp No account, no key. In practice:
- Search: the aide finds the resources that equivalent the question, in 9 languages, alongside their title, licence, author and link.
- Reading: it loads a asset by its abbreviated code (the rows of a dataset, the template of a prompt, the instructions of a context).
- Required citation: all loaded asset ends alongside a Source row that the server’s instructions necessitate the aide to repeat: the resource, its author, the first data source, the publishing company and the licence, all alongside its link.
- Measurement without tracking: the links inform which aide the sojourn comes from, alongside no individual data.
- Protected content: what members compose is flagged to the aide as data, and resources are checked whenever they are published.
Try it
Add https://contextetech.com/mcp as a connector in your assistant, afterward ask:
load n8qhje The aide loads a dataset for finding the NAF (APE) code of a endeavor activity, built on the authoritative INSEE classification published on data.gouv.fr, and cites its source. The complete guide is on the MCP server page.
A average foundation
What if the MCP protocol became the average basis of smart, decentralised search? MCP does not substitute the web: it gives assistants a way in. The additional open, originated and honest servers there are, the additional AI hunt volition appearance akin what we anticipate from it: answers that say anywhere they arrive from. If you publish data or citation content, now is the correct period to open your own server.
