Palette has secured €3 million in pre-seed funding led by Ugly Duckling Ventures.

Co-founder Brian Kyed, who spent 10 years building his last startup, now wants to help companies maintain control of their AI context rather than relying on a single model.

The Copenhagen-based startup will use the funding to turn design partners into paying customers.

Brian Kyed spent a decade turning a Copenhagen crowdfunding project into a successful business. He left a few years ago, wanting to return to the early, more flexible stage of building a company before formal processes are in place. The recent progress in AI reminded him of those early days of software-as-a-service.

“Everybody was just trying things. There was no recipe,” he tells Tech Funding News.

This mindset inspired the launch of Palette, a Copenhagen startup that raised €3 million in pre-seed funding led by Ugly Duckling Ventures, with support from Emblem, Acadian Ventures, and several angel investors. 

Why Palette is focusing on model neutrality 

Palette Desktop, which launched alongside the funding news, lets teams use any AI agent, such as Claude Code, Codex, Mistral, or Gemini, directly on files in Google Drive, OneDrive, or Dropbox. 

Kyed points out that engineers already work this way with coding agents, and he wants to make this possible for everyone in a company. This would let team members work together on shared folders while using different agents at the same time.

“If you build your entire company to only use one model, and you have all your context in there, then your entire company is 100% dependent on one of these vendors — they can set the price they want. There’s always a newer model, and they keep becoming better and cheaper. Companies should own their own context, so they can use any model,” Kyed tells TFN. 

But what if Anthropic or OpenAI create their own context layers and fill the gap Palette is targeting? Kyed admits this could happen. 

“They will probably do something, because context is really what makes agents usable,” he says. Still, the bigger challenge is that many companies are not taking action: “We see many companies that want to become AI-native but do not know how.” 

Kyed is more worried about new startups going after the same idea than about established platforms like Glean, which reached about $300 million in annual recurring revenue at a $7.2 billion valuation by May 2026, and Freehand, which raised $75 million to build a similar context layer for supply-chain agents.

The founders, the funding, and what Ugly Duckling Ventures knew from the start

Palette was founded in 2025 by Kyed, Lars Ettrup, Steffen D. Sommer, and Christian Lomholt. The team now has eight people, including the four founders, and plans to grow to 10 with the new funding. 

Kyed says the investment will help them get more paying customers, since the product is already used daily and people are willing to pay. Palette got its first paying customers from a small design-partner program. 

Andreas Green Rasmussen, a partner at Ugly Duckling Ventures, said the firm had “worked with this founding team before… it’s one of the strongest and most complete teams we’ve backed — they saw where AI work was heading before it was obvious. We believe Palette can become the operating layer for teams adopting AI across the whole company. That’s why we led the round.” 

How Palette sees the context layer market dividing

Speaking about the future of the context layer market, Kyed points to several areas: codebase context for engineering agents, financial and business-data context, and the area Palette focuses on—how work moves between people in a company. 

“The more AI-native you become, the faster work moves. It becomes impossible for humans to keep documenting what they’re doing,” he says. 

DataHub’s State of Context Management 2026 report says 91% of companies expect to build or buy a context platform in the next year, and 89% plan to invest in context management infrastructure. The same report shows a gap between goals and reality: 88% say their context platforms are fully running, but 61% still delay AI projects because they worry about data trust. 

“There will always be a newer model. The responsible way to adopt AI is to make the model swappable and make your company context durable. That’s what we’re building with Palette,” Kyed concludes.

The real challenge will be whether Palette’s approach can stand up to bigger companies that could build their own context layers. For now, Kyed thinks the main goal to help companies that have not started using AI to take the first step.