52191314/web-agent-proxy-sdk — explained in plain English
Analysis updated 2026-05-18
Let an AI agent scrape or interact with a website through a real, user-controlled browser session.
Encrypt agent browsing traffic end-to-end so a relay server cannot see the content being accessed.
Distribute browser automation requests across a pool of multiple browser runners.
Connect an AI coding assistant to the web through a Model Context Protocol server.
| 52191314/web-agent-proxy-sdk | 0xradioac7iv/tempfs | 7vignesh/pgpulse | |
|---|---|---|---|
| Stars | 0 | 0 | 0 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 4/5 | 3/5 | 4/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Needs Docker Compose or Node.js 18 plus, and a running browser runner alongside the bridge server.
Web Agent Proxy, or WAP, is infrastructure that lets AI agents browse and interact with real web pages through browser sessions that a person controls, rather than giving the agent direct, unrestricted internet access. It works by routing an agent's requests through a bridge server, which then forwards them over a connection to a registered browser runner, such as a Playwright browser instance, that actually loads the page and sends the result back. The project offers several pieces working together. There is a bridge server that coordinates traffic between agents and browser runners, small SDKs for JavaScript and Python that make it easy to send requests from an agent's code, and a browser runner component built on Playwright that performs the actual page visits. Optional end-to-end encryption can scramble the request and response content so that even the bridge server relaying the traffic cannot read the URLs, cookies, or page content being accessed. The system can also manage a pool of multiple browser runners, spreading requests across them and queuing requests when all runners are busy. It includes a real-time web dashboard for watching traffic and browser status, optional API key authentication, and the ability to capture full-page screenshots alongside scraped content. It also ships a Model Context Protocol server so tools like Claude Desktop or Cursor can use it directly, plus ready-made adapters for the LangChain and LlamaIndex agent frameworks, and a benchmark tool for testing how well agent browsing setups work against different websites. The README is explicit about intended use, stating the tool is meant for legitimate, consented automation on sites and accounts the user is authorized to access, and that it should not be used to bypass access controls, solve CAPTCHAs automatically, scrape data from services that forbid it, or access systems without permission. It can be run quickly with Docker Compose, or set up manually using Node.js for development. The README does not state a software license.
Infrastructure that routes AI agent web requests through real, consented browser sessions, with optional end-to-end encryption.
Mainly TypeScript. The stack also includes TypeScript, Playwright, Docker.
The README does not state a license, so check with the repository owner before reusing this work.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.