1. Conduid
  2. Gaming
  3. Operative WebEvalAgent
MCP server · Gaming

Operative WebEvalAgent

An MCP server to test, debug, and fix web applications autonomously.

64Good

Scored 6 months ago · breakdown

About Operative WebEvalAgent

Operative WebEvalAgent is an MCP server in the Gaming category: an MCP server to test, debug, and fix web applications autonomously.

Install

From source
git clone https://github.com/Operative-Sh/web-eval-agent

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Tools (2)

setup_browser_state🔒 Opens an interactive (non-headless) browser so you can sign in once; the saved cookies/local-storage are reused by subsequent `web_eval_agent` runs.

Parameters not published.

web_eval_agent🤖 Automated UX evaluator that drives the browser, captures screenshots, console & network logs, and returns a rich UX report.

Parameters not published.

Ask AI

✦

Ask AI about Operative WebEvalAgent

Powered by Claude · Grounded in docs

I know everything about Operative WebEvalAgent. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

  • ✓README presentThe repository ships documentation.

5 checks have not been run yet.

README

⚠️ PROJECT HAS BEEN SUNSET ⚠️

This project has been discontinued. We're building something new at withrefresh.com


🚀 operative.sh web-eval-agent MCP Server

Let the coding agent debug itself, you've got better things to do.

Demo

🔥 Supercharge Your Debugging

operative.sh's MCP Server launches a browser-use powered agent to autonomously execute and debug web apps directly in your code editor.

⚡ Features

  • 🌐 Navigate your webapp using BrowserUse (2x faster with operative backend)
  • 📊 Capture network traffic - requests are intelligently filtered and returned into the context window
  • 🚨 Collect console errors - captures logs & errors
  • 🤖 Autonomous debugging - the Cursor agent calls the web QA agent mcp server to test if the code it wrote works as epected end-to-end.

🧰 MCP Tool Reference

Tool Purpose
web_eval_agent 🤖 Automated UX evaluator that drives the browser, captures screenshots, console & network logs, and returns a rich UX report.
setup_browser_state 🔒 Opens an interactive (non-headless) browser so you can sign in once; the saved cookies/local-storage are reused by subsequent web_eval_agent runs.

Key arguments

  • web_eval_agent

    • url (required) – address of the running app (e.g. http://localhost:3000)
    • task (required) – natural-language description of what to test ("run through the signup flow and note any UX issues")
    • headless_browser (optional, default false) – set to true to hide the browser window
  • setup_browser_state

    • url (optional) – page to open first (handy to land directly on a login screen)

You can trigger these tools straight from your IDE chat, for example:

Evaluate my app at http://localhost:3000 – run web_eval_agent with the task "Try the full signup flow and report UX issues".

🏁 Quick Start

Easy Setup with One-Click Integration

  1. Get your API key (free) - when you create your API key, you'll see:
    • "Add to Cursor" button with a deeplink for instant Cursor installation
    • Prefilled Claude Code command with your API key automatically included

Manual Setup (macOS/Linux)

  1. Pre-requisites (typically not needed):
  • brew: /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
  • npm: (brew install npm)
  • jq: brew install jq
  1. Run the installer after getting an api key (free)
curl -LSf https://operative.sh/install.sh -o install.sh && bash install.sh && rm install.sh
  1. Visit your favorite IDE and restart to apply the changes
  2. Send a prompt in chat mode to call the web eval agent tool! e.g.
Test my app on http://localhost:3000. Use web-eval-agent.

🛠️ Manual Installation

  1. Get your API key at operative.sh/mcp
  2. Install uv
curl -LsSf https://astral.sh/uv/install.sh | sh
  1. Source environment variables after installing UV

Mac

source ~/.zshrc

Linux

source ~/.bashrc 
  1. Install playwright:
npm install -g chromium playwright && uvx --with playwright playwright install --with-deps
  1. Add below JSON to your relevant code editor with api key
  2. Restart your code editor

🔃 Updating

  • uv cache clean
  • refresh MCP server
    "web-eval-agent": {
      "command": "uvx",
      "args": [
        "--refresh-package",
        "webEvalAgent",
        "--from",
        "git+https://github.com/Operative-Sh/web-eval-agent.git",
        "webEvalAgent"
      ],
      "env": {
        "OPERATIVE_API_KEY": "<YOUR_KEY>"
      }
    }

Operative Discord Server

🛠️ Manual Installation (Mac + Cursor/Cline/Windsurf)

  1. Get your API key at operative.sh/mcp
  2. Install uv
curl -LsSf https://astral.sh/uv/install.sh | sh)
  1. Install playwright:
npm install -g chromium playwright && uvx --with playwright playwright install --with-deps
  1. Add below JSON to your relevant code editor with api key
  2. Restart your code editor

Manual Installation (Windows + Cursor/Cline/Windsurf)

We're refining this, please open an issue if you have any issues!

  1. Do all this in your code editor terminal
  2. curl -LSf https://operative.sh/install.sh -o install.sh && bash install.sh && rm install.sh
  3. Get your API key at operative.sh/mcp
  4. Install uv (curl -LsSf https://astral.sh/uv/install.sh | sh)
  5. uvx --from git+https://github.com/Operative-Sh/web-eval-agent.git playwright install
  6. Restart code editor

🚨 Issues

  • Updates aren't being received in code editors, update or reinstall for latest version: Run uv cache clean for latest
  • Any issues feel free to open an Issue on this repo or in the discord!
  • 5/5 - static apps without changes weren't screencasting, fixed! uv clean + restart to get fix

Changelog

  • 4/29 - Agent overlay update - pause/play/stop agent run in the browser

📋 Example MCP Server Output Report

📊 Web Evaluation Report for http://localhost:5173 complete!
📝 Task: Test the API-key deletion flow by navigating to the API Keys section, deleting a key, and judging the UX.

🔍 Agent Steps
  📍 1. Navigate → http://localhost:5173
  📍 2. Click     "Login"        (button index 2)
  📍 3. Click     "API Keys"     (button index 4)
  📍 4. Click     "Create Key"   (button index 9)
  📍 5. Type      "Test API Key" (input index 2)
  📍 6. Click     "Done"         (button index 3)
  📍 7. Click     "Delete"       (button index 10)
  📍 8. Click     "Delete"       (confirm index 3)
🏁 Flow tested successfully – UX felt smooth and intuitive.

🖥️ Console Logs (10)
  1. [debug] [vite] connecting…
  2. [debug] [vite] connected.
  3. [info]  Download the React DevTools …
     …

🌐 Network Requests (10)
  1. GET /src/pages/SleepingMasks.tsx                   304
  2. GET /src/pages/MCPRegistryRegistry.tsx             304
     …

⏱️ Chronological Timeline
  01:16:23.293 🖥️ Console [debug] [vite] connecting…
  01:16:23.303 🖥️ Console [debug] [vite] connected.
  01:16:23.312 ➡️ GET /src/pages/SleepingMasks.tsx
  01:16:23.318 ⬅️ 304 /src/pages/SleepingMasks.tsx
     …
  01:17:45.038 🤖 🏁 Flow finished – deletion verified
  01:17:47.038 🤖 📋 Conclusion repeated above
👁️  See the "Operative Control Center" dashboard for live logs.

Star History

Star History Chart


Built with <3 @ operative.sh

README copy captured when this server was indexed; it may lag the repository. The original is authoritative.

Questions

About Operative WebEvalAgent

How do I install Operative WebEvalAgent?

No package has been published for this server, so install it from source: clone the repository and follow its README.

Is Operative WebEvalAgent safe to use with an AI agent?

Its trust score is 64 out of 100 (good). It passes 1 of the 1 static security checks run so far. It has no ConduID identity yet, so agent calls to it are not receipted.

Is Operative WebEvalAgent still maintained?

Conduid hasn't recorded a commit date for this repository yet. Check the repository directly for recent activity.