Lovable absorbs Nalvin's brains
to join the fight for agent talent
Lovable, the Swedish "vibe coding" startup, has acquired the team behind Stockholm-based AI agent company Nalvin. A roadmap once built almost entirely around UI generation now gains a new axis: autonomous agents.
Nalvin's team is
joining Lovable
Lovable acquired the team behind Stockholm-based AI agent startup Nalvin, Tech.eu reported on July 31, 2026. Nalvin, founded in 2023 under the earlier name "Version Lens," was built by co-founders Pontus Gifvas, Fredrik Stockman, and Pascal Chatterjee. The startup had developed an AI agent platform for automating repetitive business tasks, and in 2024 raised a €1.5m pre-seed round from People Ventures, Curiosity, and Pitchdrive.
Deal terms were not disclosed, consistent with a classic "acqui-hire." On its own announcement page, nalvin.com frames the move as the team "continuing to build the last piece of software, at a much larger scale" inside Lovable — language that signals this is a founder-team merger, not just a technology purchase.
Keep building “the last piece of software,”
just at a much larger scale.
Four numbers behind
the deal
Against Nalvin's modest €1.5m raise, the buyer is reportedly in talks to double its valuation from $6.6bn to $13.2bn. That gap in scale underscores that this deal is about acquiring people, not technology.
From UI generation
to agents, this year's pivot
Until last month, UI generation was Lovable's main battleground — now it's steering hard into agent capability.
Lovable grew fast on vibe coding — building app UIs from a prompt — but this deal signals that alone won't sustain the lead. TechCrunch reported in March 2026 that Lovable, through M&A lead Théo Daniellot, was actively hunting for startups under a "talent before product" strategy. Nalvin marks the second publicly known acquisition, following the November 2025 purchase of cloud infrastructure provider Molnett.
| Before (UI-generation focus) | Now (agent buildout) |
|---|---|
| Generate UI/apps from a prompt | Autonomous agents that run repetitive work after generation |
| Molnett buy brought infra in-house (2025.11) | Nalvin buy brings in agent talent (2026.07) |
| Competes with generation tools like Replit, Cursor | Competition widens to autonomous-execution tools |
Who benefits, and how
The biggest impact lands on people who don't want Lovable-built apps to stop at "shipped."
Business leaders
There's a growing chance that ongoing operation, maintenance, and improvement of Lovable-built apps can be handed to agents instead of extra headcount. That's a tailwind for leaders who want automation past the initial build, without growing the dev team.
Product managers
If Nalvin's repetitive-task automation gets folded into Lovable, PMs may be able to offload post-spec work — QA passes, documentation updates, internal support triage — to agents. Worth watching for the roadmap announcement before committing process changes.
What happens next
Lovable has not yet published a concrete integration roadmap for how Nalvin's technology folds into its product. The thing to watch is how granular the agent features are when they surface in the next quarterly product announcements. For individuals and small teams doing straightforward UI generation, near-term impact is limited — the UI-generation track that was the main battleground until last month isn't going away, an agent layer is more likely being stacked on top of it. Until an integration is announced, there's no need to rush changes to existing workflows.
The risk optimism glosses over
Acqui-hires carry a well-known weakness: integration success is hard to see from the outside. Both the deal terms and the integration timeline remain undisclosed, and it's unclear how many engineers beyond Nalvin's three co-founders actually joined. Lovable also framed the November 2025 Molnett acquisition around bringing infrastructure in-house, but there's little public evidence yet of that showing up in the product. Until "agent buildout" becomes a shipped feature rather than a stated intent, it's worth treating this as a roadmap signal, not a finished capability.