Loading ToolBench
UIMax MCP — Turns Claude Code into a frontend expert that reviews and fixes your UI automatically. Free for Pro plan users.
UIMax MCP has 35 tools with significant definition gaps. Most tools lack input parameter descriptions (critical per pattern:tool-description). Descriptions are present but often verbose and marketing-focused rather than LLM-optimized (target 50-200 chars; many exceed 300). Parameter schemas exist but lack descriptions for most inputs. No output schemas documented. Error handling is absent, tools return raw results without recovery guidance. Tool naming is reasonable (verb_noun pattern mostly followed) but composition issues exist: review_ui, run_full_review, and crawl_and_review combine multiple concerns. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Baseline: 549 production tools average 194 chars descriptions; UIMax descriptions average ~250 chars and lack actionable context.
Perform static code analysis on a frontend codebase to identify quality issues, patterns, and framework usage
Navigate to a URL and capture all console messages during page load. Also captures uncaught exceptions via pageerror events.
Navigate to a URL and capture all network requests during page load. Records request/response details including status, size, and timing.
Navigate to a URL and capture all JavaScript errors, unhandled rejections, and failed resource loads.
Capture screenshots at multiple viewport sizes (mobile, tablet, desktop) for responsive design testing
Capture a screenshot of a webpage at a specific viewport size
Capture before/after screenshots and perform semantic visual comparison to verify UI changes match intent
Missing parameter descriptions on most tools. Parameters like 'url', 'codeDirectory', 'width', 'height' lack descriptions explaining format, constraints, or examples. LLMs cannot infer parameter meaning from names alone.
No output schemas documented. Tools return complex audit data (accessibility violations, performance metrics, Lighthouse scores) but LLMs have no schema to understand structure. Forces LLMs to parse unstructured responses.
Tool composition violations: review_ui, run_full_review, and crawl_and_review combine multiple concerns (screenshot + audit + analysis). Should split into single-responsibility tools so agents compose as needed.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | 2026-07-28+ | v2 |
Check if current audit results meet performance budget thresholds defined in .uimaxrc.json
Check if a webpage properly supports dark mode and analyze dark mode implementation
Click an element on the page by CSS selector
Compare two screenshots and identify visual differences using pixel-level analysis
Compare two different websites side-by-side and identify design/UX differences
Compare current audit results to a saved baseline and identify improvements or regressions
Crawl a website and run UI reviews on multiple pages to identify patterns and site-wide issues
Compare two reviews and identify changes in metrics and findings
Generate an interactive HTML report from audit results
Get information about an element on the page (text, attributes, computed styles)
Get statistics about review history (total reviews, average scores, trends)
Load the most recent baseline snapshot for a given URL
Load all review history entries from the history file
Measure Core Web Vitals and performance metrics for a webpage
Navigate to a URL in a browser page
THE PRIMARY TOOL — Fully automated UI review pipeline. Captures a screenshot, runs accessibility/performance/code audits, then returns ALL data along with an expert frontend review methodology so you can generate a comprehensive review and implement fixes. Use this when the user asks to "review my UI", "audit my frontend", or "find UI issues". After receiving the results, you MUST: 1. Study the screenshot carefully for visual/UX issues 2. Analyze the audit data following the expert methodology provided 3. Generate a comprehensive review with SPECIFIC fixes (exact CSS values, code changes) 4. Implement the fixes directly in the codebase This tool is FREE — it runs entirely within Claude Code using the user's existing plan. No API keys needed.
Run an accessibility audit on a webpage using axe-core. Injects axe-core into the page and runs a full audit, returning WCAG 2.1 violations with impact levels and fix suggestions.
Run a comprehensive UI review pipeline combining accessibility, performance, code analysis, and Lighthouse audits
Run Lighthouse audit on a webpage and return performance, accessibility, best practices, and SEO scores
Run an extended Lighthouse audit with deep analysis of PWA readiness, security, unused code, LCP optimization, and resource analysis
Run an SEO audit on a webpage checking for common SEO issues and best practices
Save a baseline snapshot of audit metrics for a URL to track changes over time
Save a review entry to the review history file
Scroll the page to a specific position or element
Select an option from a dropdown by CSS selector
Type text into an input element by CSS selector
Verify that UI fixes have been properly applied by comparing before/after screenshots and audit results
Wait for an element to appear on the page
No error handling or recovery guidance. Tools return raw results without categorizing errors as retryable, user-fixable, or fatal. No actionable error messages to guide LLM recovery.
Missing tool annotations. Tools marked as WRITE (save_baseline, click_element, type_into_element, select_option, save_review_entry) lack destructiveHint/idempotentHint annotations. READ_ONLY tools lack readOnlyHint.