Why is mobile MCP agent automation fragmented?

At least four competing mobile MCP projects exist (mobile-next, runablehq, system-pclub, CursorTouch) with no shared protocol, no interoperability, and no standard accessibility snapshot format.

What does the MobileWorld ACL 2026 benchmark show?

MobileWorld built a benchmark for autonomous mobile agents in MCP-augmented environments specifically because no standard existed. The benchmark paper itself is evidence of the gap.

How many stars does the top mobile MCP server have?

mobile-next/mobile-mcp has 5,195 GitHub stars with 450 forks, making it the dominant project despite the fragmentation.

What product could be built here?

A unified mobile MCP server that abstracts iOS (XCUITest) and Android (UIAutomator2) into a single tool schema: one MCP client connects and controls both platforms with the same tool calls. Ship the SDK, open-source the protocol.

← Back to dashboard

clawsmith.com/signal/mobile-mcp-no-stable-cross-platform-agent-automation-protocol

📈 TrendsWide OpenLive

No stable cross-platform MCP protocol for AI agent mobile automation on iOS and Android

AI agents that can control mobile apps are in high demand but there are at least four competing mobile MCP projects (mobile-next/mobile-mcp 5195 stars, runablehq/mobile-mcp, system-pclub/mobile-mcp, CursorTouch/Android-MCP) with no interoperability. MobileWorld (ACL 2026) built a benchmark specifically because no standard existed. The mobile MCP ecosystem mirrors the early MCP fragmentation at the protocol level.