CBrowser is a cognitive browser automation platform. It has 120 MCP tools that simulate how real humans think, struggle, and give up on websites.

The Problem It Solves

Traditional testing tools check if buttons work. CBrowser checks if people can use them.

A button can pass every automated test and still be unusable for someone with tremors. A checkout flow can have zero bugs and still lose 73% of first-time visitors to decision overload. Analytics tells you users leave. CBrowser tells you why.

Who Uses CBrowser

UX and Product teams use it to find where users struggle before they complain. They run cognitive journeys with different personas and see exactly which page, which element, and which cognitive layer causes abandonment.

Accessibility teams use it to go beyond WCAG checklists. Instead of checking if an aria-label exists, CBrowser simulates a user with motor tremors trying to click a 32px button on mobile. It reports the actual barrier, not just the rule violation.

Engineering leads use it to make their sites AI-agent ready. Claude, ChatGPT, and Perplexity browse the web now. CBrowser grades how well AI agents can navigate your site (A through F) and generates code patches to fix issues.

Content strategists use it to measure readability. It scores information density, sentence complexity, technical jargon, and lexical diversity per page. You see exactly which pages are too hard for your audience.

How It Works

CBrowser runs inside Claude.ai as an MCP server. You type a natural language request like "test my site's accessibility with an elderly user on mobile" and CBrowser:

Launches a real browser (Chromium)
Navigates to your site
Simulates the persona's cognitive traits (26 traits like patience, working memory, risk tolerance)
Detects barriers, measures cognitive load, and scores the experience
Returns actionable results with fix recommendations

You can also run CBrowser from the CLI (npx cbrowser) or self-host the MCP server.

How to Add CBrowser to Claude.ai

Go to Customize → Connectors → "Add custom connector"
Paste: https://demo.cbrowser.ai/mcp
Optionally, download the Claude.ai Skill (.skill) and upload it via Customize → Skills for the best experience

Claude.ai Skill vs Claude Code Skill: The Claude.ai Skill is a lightweight knowledge package for the web interface. The Claude Code Skill is a full CLI integration for the Claude Code terminal app. See Claude-Skill-Installation for both.

Core Concepts

Personas

Every user thinks differently. CBrowser models this with 26 cognitive traits per persona. A first-timer has low site familiarity and high curiosity. An impatient mobile user has low patience and high satisficing. An elderly user with low vision has reduced motor precision and increased reading tendency.

There are 21 built-in personas and you can create custom ones from a natural language description.

Cognitive Transport Chain (CTC)

The CTC measures the total cognitive cost of using a page. It has 6 sequential layers:

Saliency — Can the user find important elements?
Cognitive Load — How much mental effort does the page demand?
Decision Complexity — How many choices must the user make?
Motor Accessibility — How precisely must the user click or tap?
Frustration — How does the page handle errors and confusion?
Readability — How complex is the text?

Each layer depletes cognitive capacity for the next. The bottleneck layer is the one to fix first.

Scores

CBrowser produces four main scores:

Agent-Ready Score (0-100, grade A-F): How well AI agents can navigate the site. Categories: findability, stability, accessibility, semantics.
Empathy Score (0-100): How many barriers a specific persona encounters. Deductions from touch targets, contrast, cognitive load, timing, and sensory issues.
CTC Total Cost (0-2+): Total cognitive transport cost. Under 0.5 is easy. Over 1.0 is high demand.
Abandonment Risk (0-100%): Probability the persona gives up. Under 30% is low risk. Over 60% is high risk.

Schwartz Values

Beyond cognitive traits, personas can have motivational values based on Schwartz's Theory of Basic Human Values (10 values: achievement, benevolence, conformity, hedonism, power, security, self-direction, stimulation, tradition, universalism). These modulate how the persona responds to content — a high-achievement user pushes through friction, while a high-security user abandons at the first trust signal failure.

Main Workflows

1. Site Health Check

Run site_cognitive_assessment on your homepage. Get gate 1 (bot detection), gate 2 (agent-readiness score), and gate 3 (per-persona CTC analysis) in one call.

2. Accessibility Audit

Run empathy_audit with a disability persona (motor-impairment-tremor, low-vision-magnified, cognitive-adhd) on mobile or desktop. Get barrier list, WCAG violations, empathy score, and remediation steps.

3. Competitive Benchmark

Run competitive_benchmark with the same persona and goal on your site and 2-3 competitors. See who wins on friction, cognitive effort, and goal completion.

4. AI Readiness

Run agent_ready_audit to grade your site A-F. Then remediation_patches to generate code fixes and llms_txt_generate to create an AI-readable site manifest.

5. Content Readability

Run site_cognitive_assessment and look at the pageMetrics: information density, sentence length, technical density, lexical diversity. Compare across pages to find the hardest-to-read content.

6. Bug Detection

Run hunt_bugs to crawl your site finding console errors, broken links, accessibility issues, and visual anomalies. Then chaos_test to see how your site handles failures.

7. Custom Persona Testing

Create a persona from a description ("a busy parent checking the site on their phone during a commute") using persona_create. Then run cognitive_effort to measure how that specific user type experiences your site.

Tool Categories

CBrowser's 120 tools are organized into these categories:

Browser Automation (22 tools): Navigate, click, fill, scroll, hover, drag, type, screenshot, extract, manage tabs/cookies/storage
Cognitive Transport (7 tools): Cognitive effort, distance, coverage, interpolation, load estimation, persona comparison
Visual & Optimal Transport (11 tools): Attention heatmaps, transport maps, visual baselines, regression, cross-browser testing
Accessibility (3 tools): Empathy audit, agent-ready audit, WebMCP audit
AI Friendliness (6 tools): Agent-ready audit, benchmark, remediation patches, llms.txt, structured data
Site Knowledge (7 tools): Page understanding, site models, question answering
Testing & Quality (8 tools): Hunt bugs, chaos test, visual regression, flaky test detection
Personas (7 tools): Create, questionnaire, trait lookup, values
Marketing (5 tools): Campaign creation, execution, reporting, influence analysis
Performance (3 tools): Baselines, regression detection
Sessions (4 tools): Save, load, list, delete browser state
Security (1 tool): Security audit with tool injection scanning

Pricing

Free: 38 tools, 1 site, 50 credits/month, viewport-only analysis
Pro ($49.99/month): 97 tools, 1 site, 2,000 credits/month, full-page scope, device emulation, analytics dashboard
Enterprise: 120 tools, 10 sites, 10,000 credits/month, geo proxy, SSO, marketing suite

What Is CBrowser