We are stopping Quivr. We are building The Vibe Company.

We are stopping Quivr as our main company direction.

We are now building The Vibe Company.

This message comes from both of us, Stan and Antoine. It is not easy to write, because Quivr has been such a big part of our lives. But the reason is simple: we believe the next big opportunity is no longer to sell an isolated AI tool. It is to rebuild how companies work with agents, context, and complete workflows.

Quivr brought us here. The Vibe Company is the logical next chapter.

The previous pivot was the right move

In 2025, we wrote about how we pivoted using the YC playbook.

At the time, Quivr was coming out of a very intense period. We had an open-source product that had taken off, a large community, YC, customers, a team, and a lot of technical conviction around RAG.

But we also had a classic problem: too many use cases, too many different requests, not enough repetition. One customer wanted to analyze thousands of PDFs. Another wanted web search. Another wanted charts from CSV files. Every contract pulled us in a different direction.

Then we saw a clear signal: some customers were using Quivr much more than everyone else inside Zendesk. They were not asking for "an AI chatbot." They wanted to answer customers better and faster.

So we did what a startup has to do when a signal is stronger than its plan: we pivoted to customer support.

That pivot taught us a lot. We talked to users every week. We shipped fast. We saw what it actually means to put AI inside a real job, with constraints around quality, tone, existing tools, and trust.

It was not a mistake.

But it is no longer the right terrain for us.

What changed

Quivr reached around EUR 10k in MRR.

That is both little and a lot. Little for the SaaS ambition we had. A lot because every euro was earned in a demanding market, with real customers, real tickets, and real support teams depending on the product.

The problem was not that nobody wanted Quivr. The problem was that we could not see how to break through that ceiling fast enough, or how to build a very strong company on that specific layer.

At the same time, the market accelerated.

Autonomous agents are improving very quickly. Support platforms will keep integrating more AI features natively. Existing platforms will absorb part of the market. New agentic layers will make many specialized products harder to defend.

Our conviction became clear: standalone customer support tooling will commoditize much faster than expected.

We could keep fighting that wave.

Or we could accept what Quivr had taught us and move one layer up.

We chose the second option.

The hard decision

On April 10, 2026, we made the decision official: stop Quivr as our main direction and pivot to The Vibe Company.

That also means going back to a much smaller, much leaner, much more AI-first structure: Stan and Antoine.

This is the hardest part. A company is never just a product. It is customers, a team, investors, friends, and people who believed in a story before it was obvious.

We are grateful to everyone who built Quivr with us. The customers who tried the early versions. The open-source contributors. The investors. The team. The users who wrote to us, challenged us, and encouraged us.

None of that is lost.

But if we are honest, we were no longer building the right thing with the right company shape.

What we are building now

The Vibe Company is our attempt to build the kind of company we actually want to operate in this new AI wave.

The idea is simple:

• automate as much of our own company as possible;
• help our clients automate their real workflows;
• turn the best internal workflows into reusable tools over time.

It is not a classic agency.

It is not a classic SaaS company either.

It is an AI-native company where client delivery, internal tooling, accumulated knowledge, and future productization reinforce each other.

When we work with a client, we do not want to simply add an AI feature to their stack. We want to understand how work actually moves: documents, decisions, approvals, interfaces, exceptions, team habits. Then we build the system that allows humans and agents to execute that workflow reliably.

Quivr was a knowledge and support product.

The Vibe Company is a system for turning context into execution.

Why this is the continuation of Quivr

Quivr gave us three fundamental things.

First, a very concrete understanding of RAG and its limits. We saw what works when a company wants to connect its documents to an LLM. We also saw what breaks: poorly structured sources, implicit rules, decisions living in email, business exceptions, and quality that is impossible to judge without context.

Second, Quivr taught us that value is not in the demo. It is in the integration with real work. A useful agent is not the one that answers well once in a playground. It is the one that knows what to do on Monday morning, inside the team's tools, with the constraints of the business.

Third, Quivr taught us that companies do not only lack AI. They lack systems to compound what they know, what they decide, and what they repeat.

That is exactly what we are building with The Vibe Company.

The early signals

The pivot is recent, but the first signals are stronger than what we were seeing toward the end of Quivr.

We signed our first major client, Monka, in healthcare. In two weeks, the new direction generated more commercial traction than the previous three years of Quivr.

For Monka, we are shipping a web and mobile product inside an HDS-compliant healthcare environment in 9 weeks. A few years ago, the same scope — security posture, infrastructure, two clients, integrations — would have taken well over a year. AI-assisted engineering, a focused team, and our own internal tooling compress that timeline without compromising compliance.

Healthcare is a field we care about a lot. AI is obviously useful there, but adoption is slowed down by security, compliance, integrations, and operational complexity. That is exactly the kind of context where "just add a chatbot" is not enough.

We also have Project Mediagen with AFP, BPI, and France 2030: a long-term R&D track that lets us keep investing in the automation layer behind our work without relying only on classic SaaS revenue.

With a smaller team, lower burn, and more internal automation, we became cash-flow positive for the first time.

This is not an ending. It is a reset.

But it is a reset with traction, cash, and, most importantly, a very different kind of energy.

What changes in how we work

We no longer want to build isolated pieces.

An isolated chatbot. An isolated search tool. An isolated automation.

The new unit of value, for us, is the complete workflow.

A workflow has context. Rules. Exceptions. People who need to approve things. Existing tools. Imperfect data. Security constraints. Past decisions. A definition of good work that is almost never written down anywhere.

That is where agents become truly useful.

Not when you only give them more tools.

When you give them a method, memory, an execution environment, and a real scope of responsibility.

That is also why we are building our own internal systems, like Granite, agent skills, and our publishing, delivery, and knowledge-management pipelines. We want to be our own first client. If our own company does not run better with our tools, we have no legitimacy selling them.

We have started documenting this vision in other articles, especially on company memory for agents, our personal OS, and MCP servers as working methods.

Thank you

Quivr changed our lives.

It brought us to YC. It helped us meet thousands of developers, customers, investors, and partners. It taught us how to build under pressure. It gave us technical credibility. It also showed us, sometimes brutally, what is not enough to build a durable company.

We do not disown any of it.

But we are turning the page.

If you were a Quivr customer, contributor, investor, user, team member, or simply someone who followed the journey: thank you.

The next chapter is called The Vibe Company.

We will keep building in public. We will keep sharing what we learn. And we will put all the energy we had in Quivr behind a broader idea: helping companies work with AI as if it were truly part of their operating system.

Stan & Antoine