Skip to main content
Glama
WebEvalAgent MCP Server

WebEvalAgent MCP Server

Official

⚠️ PROJECT HAS BEEN SUNSET ⚠️

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


πŸš€ operative.sh web-eval-agent MCP Server

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

Demo

Related MCP server: agnt

πŸ”₯ 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)

  • πŸ“Š Capture network traffic - requests are intelligently filtered and returned into the context window

  • 🚨 Collect console errors - captures logs & errors

  • πŸ€– 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

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

setup_browser_state

πŸ”’ 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.

πŸ› οΈ 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

πŸ”ƒ 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

πŸ› οΈ 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

πŸ“‹ Example MCP Server Output Report

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

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

πŸ–₯️ 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 πŸ–₯️ Console [debug] [vite] connecting…
  01:16:23.303 πŸ–₯️ Console [debug] [vite] connected.
  01:16:23.312 ➑️ GET /src/pages/SleepingMasks.tsx
  01:16:23.318 ⬅️ 304 /src/pages/SleepingMasks.tsx
     …
  01:17:45.038 πŸ€– 🏁 Flow finished – deletion verified
  01:17:47.038 πŸ€– πŸ“‹ Conclusion repeated above
πŸ‘οΈ  See the "Operative Control Center" dashboard for live logs.

Star History

Star History Chart


Built with <3 @ operative.sh

Available Tools

2 tools
setup_browser_stateA

Sets up and saves browser state for future use.

This tool should only be called in one scenario:

  1. The user explicitly requests to set up browser state/authentication

Launches a non-headless browser for user interaction, allows login/authentication, and saves the browser state (cookies, local storage, etc.) to a local file.

Args: url: Optional URL to navigate to upon opening the browser. ctx: The MCP context (used for progress reporting, not directly here).

Returns: list[TextContent]: Confirmation of state saving or error messages.

Inputs: url.

web_eval_agentA

Evaluate the user experience / interface of a web application.

This tool allows the AI to assess the quality of user experience and interface design of a web application by performing specific tasks and analyzing the interaction flow.

Before this tool is used, the web application should already be running locally on a port.

Args: url: Required. The localhost URL of the web application to evaluate, including the port number. Example: http://localhost:3000, http://localhost:8080, http://localhost:4200, http://localhost:5173, etc. Try to avoid using the path segments of the URL, and instead use the root URL. task: Required. The specific UX/UI aspect to test (e.g., "test the checkout flow", "evaluate the navigation menu usability", "check form validation feedback") Be as detailed as possible in your task description. It could be anywhere from 2 sentences to 2 paragraphs. headless_browser: Optional. Whether to hide the browser window popup during evaluation. If headless_browser is True, only the operative control center browser will show, and no popup browser will be shown.

Returns: list[list[TextContent, ImageContent]]: A detailed evaluation of the web application's UX/UI, including observations, issues found, and recommendations for improvement and screenshots of the web application during the evaluation

Inputs: url (required), task (required), headless_browser.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.9/5.0

    Scored across 2 tools

    Disambiguation5/5

    The two tools have completely distinct purposes with no overlap. setup_browser_state handles browser authentication and state saving, while web_eval_agent performs UX/UI evaluation of web applications. An agent would never confuse these tools as they serve different phases of the workflow.

    Naming Consistency5/5

    Both tools follow a consistent verb_noun pattern with snake_case naming. setup_browser_state and web_eval_agent use clear action-oriented verbs followed by descriptive nouns, creating a predictable naming convention throughout the server.

    Tool Count2/5

    With only 2 tools, this server feels severely under-equipped for its apparent purpose of web evaluation. A comprehensive web evaluation system would typically need tools for navigation, interaction simulation, accessibility testing, performance measurement, and result analysis beyond just setup and evaluation.

    Completeness2/5

    The tool surface is severely incomplete for web evaluation. While setup and evaluation are covered, there are significant gaps: no tools for configuring evaluation parameters, managing browser sessions, analyzing results programmatically, testing specific components, or handling different evaluation methodologies. The server provides only basic endpoints without supporting the full evaluation workflow.

    Maintenance

    ActivityInactive
    ResponsivenessLimited data

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables AI assistants to debug frontend applications by providing direct access to browser DevTools, React state, DOM inspection, and runtime debugging capabilities. Bridges the gap between AI and complex web applications for autonomous debugging and issue resolution.
      1
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Bridges AI coding agents with the browser to provide visual debugging, real-time error capture, screenshot capabilities, DOM inspection, and interactive wireframing through a reverse proxy with injected developer tools.
      53 npm
      22
      Apache 2.0
    • F
      license
      Not graded
      quality
      D
      maintenance
      Enables AI agents to automate and debug real Chromium browsers with capabilities like screenshots, video recording, performance analysis, visual regression testing, and OCR text extraction.
      13
      -
    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables AI coding agents to access live console logs, errors, and network requests from web applications via a local WebSocket connection, without copying data to chat.
      9 npm
      19
      MIT