Lovable moves apps from human UIs to agent-accessible capabilities via MCP

Lovable, known for its AI-powered web app creation platform, is evolving toward a future where applications expose select functions as agent-accessible "capabilities" via a hosted MCP server. These capabilities can be invoked directly by agents in ChatGPT, Claude, and other MCP-compatible clients, bypassing the need for a human user to open the app. According to Lovable CTO Fabian Hedin, this approach enables a single application to present two interfaces: a traditional human UI and an agent interface.

The company frames this as part of a broader vision for a "digital brain" for organizations—an interface that aggregates context about a company and its workflows and provides reusable capabilities that agents can orchestrate. Hedin describes the goal as ensuring "everything that you’re building can be reused in an agentic way," with the platform acting as a connective layer between capabilities and agents rather than requiring separate agents for each task.

Lovable’s product evolution has been rapid. It originated from GPT Engineer, an open-source coding tool launched in 2023, and became a commercial product in November 2024 before rebranding to Lovable the following month. During this period, users began building production applications and internal tools on the platform, including CRM systems, admin panels, and customer-support consoles. Lovable now reports that employees at nearly two-thirds of the Fortune 500 have used the platform.

On the business side, Lovable says it has surpassed a $500 million annualized revenue run rate, with more than 60 million projects created and over 900 million monthly visits to Lovable-built apps. The company closed a $400 million Series C this month, led by Menlo Ventures and the Scaleup Europe Fund managed by EQT, valuing the company at $13.3 billion. Lovable attributes its pace of change to both its own application-layer innovations and the improving state of large language models.

Security and permissioning are central to the capabilities model. Lovable uses "connectors" to interface with external systems, including an "app user connector" that preserves user identity and source-system permissions. Credentials are stored server-side in encrypted form and handled by Lovable’s connector gateway, with apps receiving short-lived keys bound to the relevant user. Hedin emphasizes separating the connection to external systems from the application code to mitigate exposure risks.

Hedin acknowledges competitors like Vercel, which introduced an internal agent called @𝚟 and allows natural-language app generation via v0. However, he positions Lovable’s wedge as "being the best place to build the capabilities that agents need," arguing that orchestrating capabilities is easier than building them reliably and correctly.