Comparison
mobilerun vs mobile-use: framework plus a real fleet.
Both are open-source agent frameworks for mobile. The difference is what runs underneath: with mobile-use you bring the devices, with mobilerun they come with it.
TL;DR
mobile-use by Minitap is an Apache-2.0 multi-agent
framework. You supply the hardware, physical iOS is not supported,
and the company now leads with autonomous QA.
mobilerun ships the fleet with the framework. Real
persistent Android and physical iPhone devices. Configure their
location for any region, pay by the minute,
scale to hundreds of concurrent devices, keep state across spin-down,
and drive it all from your own models.
| Capability | mobilerun | mobile-use (Minitap) |
|---|---|---|
| What you get | Framework + fleet, devices included | Framework, bring your own devices |
| Physical iOS | Real iPhones we own and operate | Not yet supported, simulators only |
| Android devices | Persistent Android + physical iPhones on request · any-region configuration | Your own emulators or handsets |
| Device identity | Authentic hardware, the checks real apps run & app trust checks | Whatever your hardware provides |
| Product direction | Agent infrastructure for production workloads | Autonomous mobile QA, sold against BrowserStack |
| AndroidWorld (community leaderboard) | 91.4, self-reported | 91.4, self-reported. Minitap also publishes 100% of its own |
| Open source | MIT framework, 8,000+ stars | Apache-2.0 framework, ~2,700 stars |
| Interfaces | REST API, Python/TypeScript SDKs, hosted MCP | Self-hosted framework |
| Pricing (July 2026) | $0.03/min · rentals from $5/mo · iPhone $200/mo | Solo $510/mo to Scale $2,040/mo (annual), credit based |
| Device scale | Hundreds of concurrent devices | Paid tiers cap at 2 to 8 devices |
| State on spin-down | Persists, resume with logins intact | Managed by you |
| Company | Germany (EU) | $4.1M seed reported by Forbes, Dec 2025 |
Benchmark row source: the AndroidWorld community leaderboard linked from google-research/android_world lists mobile-use at 91.4 and mobilerun at 91.4. The sheet states: "Community-submitted results. No independent verification is performed. Scores are self-reported and may be inaccurate."
Minitap additionally publishes a 100% AndroidWorld figure of its own, in its README and in arXiv:2602.07787.
mobile-use details from github.com/minitap-ai/mobile-use and minitap.ai (homepage and pricing), July 2026. The README states verbatim: "Physical iOS devices are not yet supported." Prices change, check both vendors.
Why teams choose mobilerun
- Framework plus fleet. You do not have to source, rack and babysit your own handsets.
- Real iPhones we own and operate, not simulators. Native iOS identity for app trust checks.
- Persistent Android and iOS devices with location configurable for any region.
- Genuine device signals compatible with the checks real apps run.
- Per-minute economics. $0.03/min, rentals from $5/mo, dedicated Android $80/mo, real iPhone $200/mo.
- Scales on demand. Hundreds of devices at once, no seat or device cap to negotiate.
- State persists. Spin down, resume later with logins and session state intact.
- Your own models drive it. REST API, Python/TypeScript SDK, hosted MCP.
- Built for production agent workloads, not only CI test suites.
- Social media too, on your own and your clients' authorized accounts.
- Open-source framework, MIT licensed, 8,000+ GitHub stars, 91.4 on AndroidWorld.
FAQ
mobilerun vs mobile-use, common questions.
What is the main difference between mobilerun and mobile-use?+
mobile-use is an open-source multi-agent framework (Apache-2.0) that drives mobile devices you supply yourself. mobilerun is a framework plus a fleet: persistent Android and iOS devices come with it, driven by the model of your choice over REST API, Python/TypeScript SDKs and a hosted MCP server.
Does mobile-use support physical iOS devices?+
Not yet. The mobile-use README states verbatim that physical iOS devices are not yet supported, so iOS runs on simulators with a macOS host, and Minitap's paid QA product advertises iOS as simulator-backed. mobilerun runs real iPhones we own and operate, with authentic device identity for iOS apps.
How do the two compare on the AndroidWorld benchmark?+
The AndroidWorld community leaderboard linked from google-research/android_world lists mobile-use at 91.4 and mobilerun at 91.4. That leaderboard states: "Community-submitted results. No independent verification is performed. Scores are self-reported and may be inaccurate." Minitap additionally publishes a 100% AndroidWorld figure of its own in its README and paper.
Which one fits production agent workloads?+
Minitap has moved commercially toward autonomous QA and sells against BrowserStack, with credit-based seats that cap at 2 to 8 devices. mobilerun stays agent infrastructure for production: pay as you go at $0.03/minute, hundreds of concurrent devices, state that persists across spin-down on enterprise request, and real hardware identity that passes the checks real apps run.
More comparisons: vs GeeLark · vs Multilogin · Appium alternative · all comparisons
Run your agent on a real phone.
Pay as you go from $0.03/minute, or bring your own phone for $5/month.