This project provides a standardized framework, the Web Agent Protocol (WAP), for recording and replaying user interactions within a web browser. It aims to facilitate efficient web automation and agent reusability by separating action recording from execution. The Python SDK supports collecting user interaction data via a Chrome extension, converting it into replayable action lists (exact or smart replay), and generating MCP servers for agent reuse.

How It Works

WAP utilizes a Chrome extension to capture user interactions as raw event streams. The Python SDK then processes these streams, offering two conversion methods: "exact replay" for precise action duplication and "smart replay" for condensed, goal-oriented steps. These processed actions can be replayed using the WAP-Replay protocol or converted into MCP servers, enabling any agent to leverage recorded browser operations.

Quick Start & Requirements

• Install: pip install -r requirements.txt
• Prerequisites: Python 3.11, Conda environment, OpenAI and DeepSeek API keys (set in .env file).
• Setup: Requires setting PYTHONPATH to the project root. The OTA-WAP Chrome extension needs to be set up separately.
• Data Collection: Run python action_collect_server.py to receive data from the extension.
• Links: OTA-WAP Chrome Extension, WAP Replay Tool releases

Highlighted Details

• Supports both exact replay (precise action duplication) and smart replay (condensed, goal-oriented steps).
• Can generate MCP servers for reusable browser actions by any agent.
• Offers an out-of-the-box desktop app for running replay lists.
• Records session data including event details and page HTML content.

Maintenance & Community

The project acknowledges contributions from Browser-Use, MCP (modelcontextprotocol/python-sdk), and DOMListenerExtension. Further community or roadmap information is not detailed in the README.

Licensing & Compatibility

The README does not explicitly state a license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The "exact replay" and "smart replay" generators require recorded sessions to contain exactly one task-start and one task-finish event. The ModuleNotFoundError suggests potential issues with environment setup or PYTHONPATH configuration if commands are not run from the project root.