What changed
cdpilot introduces a command-line interface for browser automation that bypasses traditional libraries such as Puppeteer, Playwright, and Selenium. It operates directly using the Chrome DevTools Protocol (CDP) over HTTP/WebSocket, aiming for zero npm dependencies on the CLI side. The Python backend requires only the websockets package. The tool provides over 70 commands, including navigation, taking screenshots, network inspection, console command execution, accessibility checks, and video understanding. It also supports progressive anti-bot resilience and test assertions. A key feature is its MCP (Meta-Command Protocol) server, designed for integration with AI coding agents like Claude, GPT, and Gemini, facilitating tool-use workflows.
cdpilot emphasizes isolated browser sessions, ensuring that personal browser profiles remain untouched. Each project gets its own automatically managed browser instance and port to prevent conflicts. For AI agents, it provides visual feedback during automation, such as a green glow overlay and click ripples, along with an AI control warning when hovering during active automation. Privacy is a focus, with all operations running locally unless explicitly opted into for features like CAPTCHA solvers or proxies.
The tool includes a workload-aware browser selection mechanism. By default, it picks the most suitable browser based on whether extensions are registered or for pure automation tasks. It intelligently selects between Vivaldi, Brave, Edge, Chromium, and Chrome (including Chrome for Testing) based on platform stability and extension compatibility, addressing issues like Chrome 137+ silently dropping support for loading unpacked extensions. For extension development, cdpilot facilitates the installation and use of Chrome for Testing (CfT), a versioned Chrome build designed for automation that still supports unpacked extensions. It manages CfT installation, extraction, and automatic selection when extensions are being developed.
Installation can be done directly via npx cdpilot or by global npm installation. Requirements include Node.js 18+, Python 3.10+, and one of Brave Browser, Google Chrome, or Chromium. The first-time setup involves npx cdpilot setup to auto-detect browsers and create isolated profiles. For AI integration, cdpilot offers a Claude Code plugin that can be installed directly from the marketplace, registering the MCP server and a skill to guide Claude Code on using cdpilot commands. The MCP server has configurable options for allowing file access (CDPILOT_MCP_ALLOW_FILE), direct browser connections (CDPILOT_MCP_ALLOW_CONNECT), and environment variable passing (CDPILOT_MCP_ALLOW_VARS), with default restrictions for security.
Why it matters for builders
cdpilot's direct CDP approach and minimal dependencies significantly reduce setup complexity for browser automation tasks. Builders can integrate powerful automation capabilities into their projects without wrestling with dependency conflicts or complex configurations. The tool's design for AI agents, coupled with its extensive command set, opens up new possibilities for creating sophisticated AI-driven workflows that interact with web applications.
Practical impact
Developers can immediately start using cdpilot via npx cdpilot for tasks ranging from simple website screenshots to complex end-to-end testing. For those working with AI agents, installing the Claude Code plugin provides a direct pathway to leverage cdpilot's browser control features within their LLM workflows. Experimenting with the 70+ commands and exploring the browser selection logic will reveal opportunities for optimizing automation scripts and agent behaviors.
Caveats and source limits
The source material describes cdpilot as a fresh release with specific GitHub metrics (33 stars, 8 forks) and mentions "5 AI signals, 6 developer signals" without providing details on these signals. The latest release version is v0.9.12. While the tool is designed for AI agents, specific benchmark results or independent performance comparisons against tools like Puppeteer or Playwright are not provided. The source also details various environment variables and MCP server configurations, but practical examples of their usage in complex AI agent scenarios are limited. The exact scope of "progressive anti-bot resilience" is not fully elaborated.
Sources
Claim check: 12/12 supported claims - 12 evidence links - 100% avg confidence
- cdpilot is a browser automation CLI that uses raw CDP without Puppeteer, Playwright, or Selenium.supported - github.com
- cdpilot has over 70 commands for navigation, test assertions, and interaction.supported - github.com
- cdpilot includes an MCP server for AI coding agents.supported - github.com
- cdpilot offers isolated browser sessions for each project.supported - github.com
- cdpilot supports AI agents like Claude, GPT, and Gemini.supported - github.com
- cdpilot provides visual feedback during automation, such as a green glow overlay.supported - github.com
- cdpilot automatically selects browsers based on workload, supporting Vivaldi, Brave, Edge, Chromium, and Chrome.supported - github.com
- cdpilot facilitates the installation and use of Chrome for Testing (CfT) for extension development.supported - github.com
- cdpilot requires Node.js 18+, Python 3.10+, and a compatible browser (Brave, Chrome, or Chromium).supported - github.com
- cdpilot offers a Claude Code plugin for direct integration with the MCP server.supported - github.com
- cdpilot's MCP server can be configured to allow file access, direct browser connections, and environment variable passing.supported - github.com
- cdpilot is a fresh release with 33 stars and 8 forks on GitHub.supported - github.com
Caveats
- Single-source caution: verify critical details at the linked source.
Radar score 88/100 - how it was calculated
- Reliability 82: GitHub metadata supports source trust
- Freshness 95: Fresh GitHub release date
- Novelty 81: Fresh GitHub release
- Technical 85: Repository technical metadata
- Developer 96: Developer tooling signals
- Ecosystem 72: Fresh GitHub release
- Confidence 100: Claims have reliable evidence
Discussion
Loading comments...