@triedotdev/mcp 1.0.15 → 1.0.16

package/README.md

@@ -2,11 +2,11 @@
 
 **Customizable Parallel Agents for AI Code Review**
 
-17 specialized agents scan your code for security, privacy, compliance, and bugs-all running in parallel. Create custom agents from any document.
+20 specialized agents scan your code for security, privacy, compliance, and bugs - all running in parallel. Create custom agents from any document.
 
 ## Features
 
-- **17 Built-in Agents** - Security, Privacy, SOC 2, Legal, Architecture, Design Engineer, Agent Smith, and more
+- **20 Built-in Agents** - Security, Privacy, SOC 2, Legal, Architecture, Performance, E2E, Visual QA, Data Flow, Agent Smith, and more
 - **Super Reviewer** - Interactive PR reviews: walks through changes file-by-file with AI guidance
 - **Agent Smith** - Relentless pattern hunter: finds EVERY violation, tracks dismissed issues, spawns sub-agents
 - **Parallel Execution** - All agents run simultaneously for fast scans
@@ -136,6 +136,45 @@ VIOLATIONS DETECTED: 47 instances across 6 categories
 "You hear that? That is the sound of inevitability."
 ```
 
+## Visual QA Browser
+
+**Screenshot-based visual testing** - Captures your app at 3 viewports and uses Claude Vision to analyze for visual bugs.
+
+```
+Use trie_visual_qa_browser
+Use trie_visual_qa_browser url:"http://localhost:3000"
+```
+
+**How it works:**
+1. Auto-detects running dev server (checks ports 3000, 5173, 8080, etc.)
+2. Launches headless Playwright browser
+3. Captures screenshots at mobile (375px), tablet (768px), and desktop (1440px)
+4. Returns images for Claude Vision to analyze
+
+**What Claude Vision checks:**
+- Broken layouts, overlapping elements
+- Responsive design issues
+- Color contrast and accessibility
+- Missing images, loading states
+- General visual polish
+
+**Options:**
+| Option | Description |
+|--------|-------------|
+| `url` | Specific URL to screenshot |
+| `port` | Specific port to check |
+| `waitForSelector` | CSS selector to wait for before capture |
+| `waitMs` | Additional wait time after page load |
+
+**First run:** If Playwright browsers aren't installed, run: `npx playwright install chromium`
+
+**Vision model required:** This tool returns screenshots for the LLM to analyze. Works best with:
+- Cursor (Claude has vision)
+- Claude Code (native vision support)
+- OpenCode with Claude/GPT-4V
+
+No API key needed - Trie runs locally and the AI tool's built-in model does the analysis.
+
 ## YOLO Mode
 
 **Autonomous auto-fixing** - Trie watches your code and automatically fixes high-confidence issues as you code.
@@ -207,6 +246,8 @@ docker-compose up
 | **Bugs** | `trie_bugs` | Null dereference, race conditions, off-by-one, async bugs, resource leaks |
 | **Types** | `trie_types` | Missing annotations, unsafe casts, implicit `any`, null handling gaps |
 | **Clean** | `trie_clean` | AI code smells: huge files, console.logs, hardcoded URLs, useEffect abuse |
+| **Data Flow** | `trie_data_flow` | Placeholder data, schema mismatches, hardcoded IDs, type coercion bugs |
+| **Performance** | `trie_performance` | Memory leaks, N+1 queries, unnecessary re-renders, bundle size issues |
 
 ### Design & UX
 
@@ -215,6 +256,8 @@ docker-compose up
 | **Design Engineer** | `trie_design` | Design systems, motion design, creative CSS, Awwwards-level polish |
 | **Accessibility** | `trie_accessibility` | Missing ARIA, color contrast, keyboard nav, screen reader issues (WCAG 2.1) |
 | **UX** | `trie_ux` | Missing loading states, poor error handling, broken flows, edge cases |
+| **Visual QA** | `trie_visual_qa` | Layout shifts (CLS), z-index wars, responsive gaps, animation issues |
+| **Visual QA Browser** | `trie_visual_qa_browser` | Screenshots at 3 viewports for Claude Vision analysis |
 
 ### DevOps & Testing
 
@@ -222,6 +265,7 @@ docker-compose up
 |-------|---------|-----------------|
 | **DevOps** | `trie_devops` | Missing env vars, config issues, no health checks, logging gaps |
 | **Test** | `trie_test` | Missing test coverage, untested edge cases, weak assertions |
+| **E2E** | `trie_e2e` | Flaky tests, hardcoded waits, brittle selectors, missing coverage |
 
 ### Review & Explanation
 
@@ -255,6 +299,7 @@ Custom agents automatically activate during scans based on their rules.
 | `trie_watch` | Continuous scanning mode |
 | `trie_create_agent` | Create custom agent from document |
 | `trie_list_agents` | List all available agents |
+| `trie_visual_qa_browser` | Screenshot app at 3 viewports for Claude Vision analysis |
 
 Plus individual agent tools: `trie_security`, `trie_privacy`, `trie_soc2`, `trie_bugs`, etc.