astrolabe
Enables AI agents to inspect and verify mobile UI hierarchies, nodes, styles, screenshots, and runtime attributes of iOS apps in simulators or on USB devices.
README
Astrolabe
Astrolabe is a runtime UI inspection tool for AI coding agents. It combines a Swift CLI, a local MCP server, and a packaged Codex skill to inspect and verify running mobile interfaces.
Features
- Discover supported apps running in iOS simulators or on paired USB devices.
- Inspect UI hierarchies, node frames, visibility, text, styles, semantic roles, and detailed runtime attributes.
- Capture native-resolution screenshots and compare them with expected images or recorded baselines.
- Query nodes and verify style, layout, and node-detail expectations.
- Keep related hierarchy operations bound to a frozen page snapshot.
- Apply allowlisted, in-memory presentation patches in Debug builds and revert them without changing source code or persistent application state.
- Expose the same inspection workflow through CLI commands, MCP tools, and the
bundled
astrolabeskill.
Requirements
- macOS 13 or later
- Swift 5.9 or later
- Node.js 22 or later
- A supported app with an Astrolabe Runtime SDK enabled in Debug builds
Installation
Clone the repository and install Astrolabe for Codex:
git clone https://github.com/regulusleow/astrolabe.git
cd astrolabe
npm run install:codex
The installer builds the CLI and MCP adapter, installs the package under
~/.astrolabe/package, links the bundled skill into
~/.agents/skills/astrolabe, and updates the managed Astrolabe sections in
~/.codex/config.toml.
Restart Codex or open a new Codex session after installation.
Installation management commands:
npm run reinstall:codex
npm run update:codex
npm run check:codex
npm run uninstall:codex
Quick Start
Launch a Debug build that includes a supported Runtime SDK, then discover it:
astrolabe list-apps --json
Use the returned appId with inspection commands:
astrolabe inspect-screen <app-id> --json
astrolabe capture-hierarchy <app-id> --json
astrolabe find-nodes <app-id> --role text --visible-only --limit 20 --json
astrolabe inspect-node <app-id> --oid <oid> --json
astrolabe node-detail <app-id> <oid> --json
astrolabe capture-screenshot <app-id> --output /tmp/screen.png --source auto --json
Hierarchy commands return a snapshotId. Pass it to subsequent hierarchy,
node, style, and layout commands to keep the workflow bound to the captured
page. Screenshot and visual comparison commands always use the latest screen.
The installed MCP server exposes the same capabilities to Codex. The bundled skill guides app discovery, snapshot reuse, node selection, runtime checks, screenshots, visual comparison, and temporary presentation experiments. Codex is currently the only officially supported AI coding platform.
Development
Install dependencies and run the test suites:
npm ci
npm --prefix mcp-adapter ci
npm test
swift test --parallel
Build the Release CLI:
swift build -c release --product astrolabe
Run end-to-end smoke tests with a supported app running:
npm run test:smoke
npm run test:usb
Repositories
- astrolabe-protocol: platform-neutral Wire Protocol, Schemas, Fixtures, and Swift DTOs.
- astrolabe-runtime-ios: Debug-only UIKit Runtime SDK.
Astrolabe Host consumes the shared protocol directly and does not depend on a platform Runtime implementation package.
Roadmap
- Android Runtime and Host support.
- Integrations for additional AI coding platforms.
License
Astrolabe is available under the Apache License 2.0. Components derived from third-party projects remain subject to their original licenses; see THIRD_PARTY_NOTICES for details.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。