# Cognitive Browser

Cognitive Browser connects AI agents to web browsers for user journey simulation, visual diffing, and accessibility audits. Useful for end-to-end regression testing.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · 19566 installs
- Author: alexandria-shai-eden
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://smithery.ai/server/alexandria-shai-eden/cbrowser
- Updated: 2026-04-12
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/mcp/cognitive-browser

## What it does

The server exposes over one hundred browser control tools for layout inspection and automated bug discovery. It finds DOM targets by intent and accessibility tree roles rather than brittle CSS paths. Built-in helpers remove cookie dialogs, run natural language test suites, and generate attention heatmaps based on distinct user profiles. Layout regression tests rely on Wasserstein consensus math to prevent false alarms on dynamic components.

Every multi-step flow relies on `_browserToken`, which is generated during initial navigation. If an agent calls `click`, `fill`, or `screenshot` without forwarding that token, the command runs against an empty new browser window instead of your current tab. Persisting cookies or local storage between test suites is handled through `save_session` and `load_session`.

The public endpoint is free for evaluation and enforces a hard rate limit of 50 requests per hour.

## Requirements
- Remote server: `https://cbrowser.run.tools`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http cbrowser https://cbrowser.run.tools
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add cognitive-browser --url https://cbrowser.run.tools
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "cognitive-browser": {
            "url": "https://cbrowser.run.tools"
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "cognitive-browser": {
            "type": "http",
            "url": "https://cbrowser.run.tools"
        }
    }
}
```

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http cognitive-browser https://cbrowser.run.tools
```

## When to use it
- track visual regressions between layout versions
- find elements by intent and accessibility markers
- simulate different user personas across application journeys
- dismiss cookie banners and modal overlays automatically

## Example prompts
- Navigate to the target site via `navigate` and locate the signup button using `find_element_by_intent`.
- Run `visual_regression` against the active baseline to identify layout drift.
- Clear cookie banners on the open page with `dismiss_overlay`.

## Similar
- [gem-browser-tester](https://affpapa.org/en/nevibecode/agents/gem-browser-tester.md) — subagent · ★ 41430 repo stars
- [Browser Automation](https://affpapa.org/en/nevibecode/skills/claude-office-skills-browser-automation.md) — skill · 5400 installs
- [browser-replay](https://affpapa.org/en/nevibecode/skills/browser-replay.md) — skill · 721 installs
- [browser-navigation](https://affpapa.org/en/nevibecode/skills/browser-navigation.md) — skill · 120 installs
- [agent-browser](https://affpapa.org/en/nevibecode/plugins/agent-browser.md) — plugin · ★ 1216 repo stars
- [browser-automation](https://affpapa.org/en/nevibecode/skills/different-ai-browser-automation.md) — skill · 1 install

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
