VoiceRun raises $5.5M to build a “voice agent factory” for developers

As companies race to add AI-powered voice agents to customer service lines, apps, and products, many developers are finding the current tools frustratingly limited. That frustration is precisely what led VoiceRun to exist.

Co-founders Nicholas Leonard and Derek Caneja originally set out to build voice agents themselves. What they found instead was a market split in two. On one side were no-code tools that made it easy to launch quickly, but often produced brittle, low-quality agents. On the other hand, there were highly customised systems that required months of engineering work and significant resources.

“There needed to be another option,” Leonard said. Around the same time, he and Caneja were also thinking more broadly about where software development is heading. Their view was that future systems would increasingly be written, tested, and improved by coding agents — and that those agents would work best by working directly in code, not through visual flowcharts.

That thinking became the foundation for VoiceRun, a platform that enables developers to build, deploy, and scale voice agents, with code as the primary interface.

Most voice agent platforms today rely on visual builders, in which developers click through conversation paths and fill in prompt boxes to define behaviour. That approach works for simple demos, but becomes difficult to manage as products grow more complex, Leonard said.

VoiceRun takes a different path. Developers define how agents behave directly in code, giving them greater flexibility and control. According to Leonard, this is also a better match for modern AI systems.

“Code is the native language of coding agents,” he said. “They can operate far more effectively there than inside a visual interface.”

That flexibility matters when teams want to go beyond basic use cases. Adjusting speech patterns, supporting regional dialects, or handling edge-case interactions can be tricky—or impossible—if a visual tool isn’t available. In code, those changes are relatively straightforward.

Alongside code-based development, VoiceRun also supports features like A/B testing and one-click deployment. The company is focused on enterprise developers building production-grade voice systems, including customer support tools and voice-first products. Leonard pointed to a restaurant technology company that uses VoiceRun to launch an AI phone concierge for reservations.

This week, VoiceRun announced it has raised a $5.5 million seed round, led by Flybridge Capital.

The funding arrives amid heavy competition in the AI agent space, which drew billions of dollars in investment last year. Leonard sees VoiceRun positioned between quick-build no-code platforms like Bland and Retell AI, and lower-level infrastructure tools such as LiveKit and Pipecat that offer maximum control but require more engineering effort.

“We provide global voice infrastructure and an evaluation-driven lifecycle, while keeping ownership of business logic and data in the customer’s hands,” Leonard said. He describes VoiceRun as closing the loop on end-to-end development, where developers supervise coding agents that write code, run tests, deploy updates, and suggest improvements.

Leonard also hopes that better tooling can help change how people perceive automated voices. Many customers still prefer human agents, often because voice automation has earned a reputation for being frustrating. A Five9 survey from last year found that about three-quarters of respondents still prefer speaking with a human for customer service.

Leonard argues that human agents have limitations as well, ranging from language barriers to interactions that can feel judgmental or inconsistent.

“There were great cars before the Model T,” he said. “But cars didn’t become ubiquitous until the assembly line. There are great voice agents today, but they won’t be everywhere until the voice agent factory exists. That’s what we’re building with VoiceRun.”