Linkup: €10M to build the search engine AI agents actually need

This is the exact problem Linkup is tackling, and the reason behind its recent €10 million funding round. The goal is not to create yet another search tool, but to build an infrastructure designed from the ground up for AI agents.

For Philippe, the gap is obvious.

“Today, agents are very good at reasoning, but very bad at accessing reliable and up-to-date information.”

Behind this statement lies a structural belief: without reliable access to external information, agents will remain impressive demos rather than systems ready for production.

Why AI agents need a new kind of search engine

Language models are trained on massive datasets over long periods of time. Once training is complete, their understanding of the world is frozen.

Philippe emphasizes this often misunderstood limitation.

“Models are trained for months, and then they no longer know what happens next.”

As soon as information becomes recent, agents hit a hard boundary. Either they admit they do not know, or they attempt to produce a plausible answer.

That is where risk emerges.

“The weaker models hallucinate and don’t tell you.”

For Linkup, this cannot be solved by better reasoning alone. It requires a dedicated layer focused on information access.

“The problem is not reasoning. The problem is access to information.”

Why traditional web search is not enough

For decades, search engines have been designed for humans. Their role is to surface links and let users do the rest.

Philippe highlights the mismatch with AI agents.

“Web search engines return lists of links. Humans can click, read, and interpret. Agents cannot.”

Agents do not need pages or websites. They need immediately usable information that can be injected into a reasoning pipeline.

“An agent doesn’t need a list of websites. It needs information.”

This gap explains why Linkup chose to rebuild search from scratch rather than adapt existing systems.

Indexing information instead of pages

Linkup’s core technical decision is radical and explains much of the product’s complexity.

Instead of indexing full web pages, the platform breaks content down into atomic units.

“We don’t index web pages. We index units of information.”

Each page is analyzed, structured, and split into blocks that agents can consume directly.

“We index atoms of information.”

This approach makes integration extremely smooth for developers.

“You can feed Linkup’s output directly into your agent.”

But it also requires rebuilding a Web index specifically for machine consumption.

“Rebuilding a web index for agents is extremely hard.”

€10M to build critical infrastructure

The €10M funding round makes sense in this context. Linkup is not financing a feature or an interface layer.

Philippe is explicit.

“We are not building a feature. We are building infrastructure.”

Continuously indexing, refreshing, and structuring the Web is capital-intensive and forces hard trade-offs.

“Indexing the Web blindly is an extremely expensive mistake.”

This is why Linkup makes deliberate choices about scope.

A search engine designed for professional use cases

Linkup does not aim for full coverage of the Internet.

Philippe explains this positioning clearly.

“We don’t need to cover the folkloric corners of the Web.”

The engine focuses on use cases where reliability truly matters: financial research, competitive intelligence, sales automation, internal assistants.

“The use cases that work are professional ones.”

This focus allows Linkup to concentrate investment where it creates the most value.

Conclusion

With its €10M funding round, Linkup is not trying to build a better consumer search engine.

Philippe sums it up simply.

“We are a tool for developers.”

The ambition is more foundational. To build the search engine AI agents need to access reliable, current, and verifiable information.

A discreet but essential infrastructure layer to move AI agents from promise to production.