io.github.hoainho/podium-mcp

Overview

An MCP server that exposes 51 tools for AI agents to automate mobile and canvas UIs. It covers iOS and Android automation, Maestro end-to-end flows, evidenced assertions, and React Native/Metro debugging. For canvas or WebGL games, it offers a no-vision approach that drives Pixi, Konva, Fabric, Phaser, Three, and Babylon UIs like DOM elements, with a claimed cost roughly 5x lower than screenshot loops.

Best for

Best for
AI agent builders who need MCP-based automation for mobile apps and canvas or WebGL games

Use cases

• Automate iOS and Android app flows with AI agents
• Debug React Native and Metro projects through MCP tools
• Drive canvas or WebGL game UIs without screenshot-based vision

How to use

Tools exposed

• tap_on
• inspect_screen
• engine_inspect
• engine_tap
• engine_swipe
• engine_call
• canvas_inspect
• canvas_resolve
• canvas_tap
• podium_token_report
• podium_health
• device_list
• device_boot
• screen_size
• orientation_get
• set_location
• open_url
• app_install
• app_launch
• app_terminate

Tested with

Claude Code

Example client config

{\n  "mcpServers": {\n    "podium": { "command": "npx", "args": ["-y", "podium-mcp"] }\n  }\n}

Notes

An MCP server that exposes 51 tools for AI agents to automate mobile and canvas UIs. It covers iOS and Android automation, Maestro end-to-end flows, evidenced assertions, and React Native/Metro debugging. For canvas or WebGL games, it offers a no-vision approach that drives Pixi, Konva, Fabric, Phaser, Three, and Babylon UIs like DOM elements, with a claimed cost roughly 5x lower than screenshot loops.

4 stars on GitHub. Last updated 2026-06-30. Licensed MIT.

Use cases

• Automate iOS and Android app flows with AI agents
• Debug React Native and Metro projects through MCP tools
• Drive canvas or WebGL game UIs without screenshot-based vision

Pros

• Broad tool coverage across mobile and canvas domains in one MCP server
• No-vision canvas interaction avoids screenshot overhead and reduces cost
• Evidenced assertions support verifiable end-to-end checks

Cons

• Low GitHub star count suggests an early-stage project with limited community traction
• E2E flows rely on Maestro, which may require separate installation and configuration
• No-vision canvas approach may miss visual states that only a screenshot would reveal

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.