Most AI agents work well on desktop and web. Mobile has always been a different story. Droidrun is a Berlin-based startup fixing exactly that, and it’s doing so with a technical approach that gets to the root of why mobile automation has been hard. CEO Christian Ninstel describes the product simply: “Like creating an API for every mobile app ever built.”
Droidrun was founded in 2024 by five co-founders: Christian Ninstel, Niels Schmidt, Nikolai Dück, Peter Lächner, and Timo Beckmann. Ninstel, Schmidt, Dück, and Lächner have built two companies together before this one, which gives them a tested working dynamic. Schmidt, who serves as CTO, built the core automation architecture. Dück, as CPO, built an AI chatbot platform seven years ago that Fortune 50 companies still use today. Lächner brings enterprise sales experience from his time at Bosch. That mix of technical depth and enterprise go-to-market experience is an unusual combination at the pre-seed stage.
Why the Accessibility API Changes Everything
Most existing tools try to make AI “see” the phone screen like a human would, essentially taking screenshots and interpreting visuals. Droidrun takes a different path. It converts mobile interfaces into structured text data that large language models can directly interpret, using Android’s Accessibility API. This approach enables precise and fast control of apps, effectively turning any mobile application into an API that AI agents can interact with consistently in real-world conditions.
The result is three times faster execution, lower computational costs, and much higher reliability when apps update their UIs, precisely the scenarios where vision-based solutions tend to fail.
Numbers That Back It Up
Droidrun’s open-source framework picked up real developer interest fast. Over 900 developers signed up within 24 hours of launch, and the GitHub repository crossed 3,400 stars within the first 10 weeks.
On the AndroidWorld benchmark, which tests AI agents across 116 diverse Android tasks, Droidrun achieved a 91.4% success rate, completing 106 out of 116 tasks using GPT-5 and direct Accessibility API access, outperforming all competing agents on the leaderboard. That’s a strong result for a framework still in its early stages.
The €2.1M Pre-Seed Round
Droidrun raised €2.1M in pre-seed funding, led by Merantix Capital, with participation from Sixty Degree Capital and angel investors including Peter Sarlin (Co-Founder of Silo AI, acquired by AMD), Philipp Langnickel, Mandeep Singh (Venture Partner at Hoxton Ventures), and Felix Jahn (Founder of McMakler and Home24).
Adrian Locher, co-founder and General Partner at Merantix Capital, put the thesis directly: “Mobile remains the dominant user interface, and the need for automation is accelerating. The problem: there is no sufficient existing solution for true agentic AI use on mobile. Until now.”
What You Can Actually Build
Droidrun fits into three practical use cases right now. The first is automating repetitive mobile tasks, such as logins, bookings, and form submissions, on real or virtual devices. The second is accessing data that only lives inside mobile apps, behind logins or rate-limited APIs. The third is integrating mobile execution into existing stacks using tools like N8N or custom scripts.
Droidrun supports multiple LLM providers, including OpenAI, Anthropic, and Gemini, while offering planning capabilities for complex multi-step tasks and self-healing automation that adapts to app changes. For developers already building AI agent workflows, this is the infrastructure layer that makes mobile part of the picture. You can explore the open-source framework on Droidrun’s GitHub to get started.
Cloud Product Is Coming
Beyond the open-source library, Droidrun is building a hosted cloud product. The cloud platform lets you run AI agents at scale across multiple devices in parallel, with credential management and complex workflow orchestration built in.
The credential management piece is particularly useful for teams deploying across multiple apps. Instead of handling app logins at the automation layer each time, users enter credentials once, and the platform manages the rest.
For founders and developers thinking about AI agent infrastructure, Droidrun represents one of the more technically grounded approaches to mobile automation available today.
