kaboom-browser-ai-devtools
Browser debugging and inspection toolkit for AI coding assistants via MCP protocol
How do I install this agent skill?
npx skills add https://github.com/reason-machines/devtools-skills --skill kaboom-browser-ai-devtoolsIs this agent skill safe to install?
- Gen Agent Trust Hubfail
This skill instructs users to install software by piping remote scripts directly from an untrusted GitHub repository into a system shell (bash/iex). This practice is highly dangerous as it allows for arbitrary code execution on the host system without prior verification. Furthermore, the skill's ability to capture and process sensitive browser data, combined with browser automation capabilities, presents a high risk of data exfiltration or system manipulation if the tool is compromised.
- Socketwarn
1 alert: gptAnomaly
- Snykfail
Risk: CRITICAL · 4 issues
What does this agent skill do?
Kaboom Browser AI DevTools MCP
Skill by ara.so — Devtools Skills collection.
Kaboom is a browser debugging, inspection, and verification toolkit that streams console logs, network failures, exceptions, recordings, and browser evidence into MCP-compatible coding assistants. It provides real-time browser debugging without requiring --remote-debugging-port, using a standard browser extension instead.
Key Features
- Console monitoring — Stream all
console.log(),.warn(),.error()with full arguments - Network debugging — Capture failed requests (4xx/5xx) with full request/response bodies
- Exception tracking — Uncaught errors with complete stack traces
- WebSocket monitoring — Connection events and message payloads
- User action recording — Click, type, navigate, scroll with smart selectors
- Web Vitals tracking — LCP, CLS, INP, FCP with regression detection
- DOM inspection — Query page elements via CSS selectors
- Accessibility audits — WCAG compliance checks with SARIF export
- Security audits — Credentials, PII, headers, cookies analysis
- Browser automation — Click, type, select, upload, navigate programmatically
- Test generation — Generate Playwright tests from recorded sessions
- Visual annotations — Drawing mode for user feedback with style extraction
Installation
Quick Install (Recommended)
macOS / Linux:
curl -sSL https://raw.githubusercontent.com/brennhill/Kaboom-Browser-AI-Devtools-MCP/STABLE/scripts/install.sh | bash
Windows (PowerShell):
irm https://raw.githubusercontent.com/brennhill/Kaboom-Browser-AI-Devtools-MCP/STABLE/scripts/install.ps1 | iex
This automatically:
- Downloads the binary for your platform
- Installs browser extension to
~/.kaboom/extension - Auto-configures detected MCP clients (Claude Code, Cursor, Windsurf, Zed)
Manual Installation
- Download binary from releases
- Extract and make executable:
tar -xzf kaboom-*.tar.gz chmod +x kaboom mv kaboom /usr/local/bin/ - Install browser extension:
- Open
chrome://extensions - Enable Developer mode
- Click "Load unpacked"
- Select
~/.kaboom/extensiondirectory
- Open
MCP Configuration
Add to your MCP client config:
Claude Desktop / Claude Code (claude_desktop_config.json):
{
"mcpServers": {
"kaboom": {
"command": "/usr/local/bin/kaboom",
"args": ["--mcp"]
}
}
}
Cursor (config.json in .cursor/ directory):
{
"mcpServers": {
"kaboom": {
"command": "/usr/local/bin/kaboom",
"args": ["--mcp"]
}
}
}
Zed (settings.json):
{
"context_servers": {
"kaboom": {
"command": {
"path": "/usr/local/bin/kaboom",
"args": ["--mcp"]
}
}
}
}
MCP Tools Reference
Console & Error Monitoring
kaboom_console_clear — Clear all captured console logs
// No arguments required
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_console_clear"
});
kaboom_console_get — Get all captured console logs
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_console_get",
arguments: {
level: "error" // Optional: "log", "warn", "error", "info"
}
});
Network Monitoring
kaboom_network_get — Get captured network requests
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_network_get",
arguments: {
status: "failed", // "all", "failed", "success"
includeBody: true
}
});
kaboom_network_clear — Clear network capture buffer
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_network_clear"
});
DOM Inspection
kaboom_dom_query — Query DOM elements with CSS selectors
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_dom_query",
arguments: {
selector: ".error-message",
includeStyles: true,
includeAttributes: true
}
});
kaboom_dom_snapshot — Capture full DOM snapshot
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_dom_snapshot",
arguments: {
includeInlineStyles: true
}
});
User Action Recording
kaboom_recording_start — Start recording user actions
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_recording_start",
arguments: {
captureClicks: true,
captureInputs: true,
captureNavigation: true,
captureScrolls: true
}
});
kaboom_recording_stop — Stop and retrieve recording
const recording = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_recording_stop"
});
// Returns array of recorded actions with selectors
kaboom_recording_generate_test — Generate Playwright test from recording
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_recording_generate_test",
arguments: {
format: "playwright", // "playwright" or "puppeteer"
includeAssertions: true
}
});
Browser Automation
kaboom_browser_click — Click element
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_click",
arguments: {
selector: "button[type='submit']"
}
});
kaboom_browser_type — Type into input field
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_type",
arguments: {
selector: "input[name='email']",
text: "user@example.com",
clear: true
}
});
kaboom_browser_navigate — Navigate to URL
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_navigate",
arguments: {
url: "https://example.com/dashboard",
waitForLoad: true
}
});
kaboom_browser_select — Select dropdown option
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_select",
arguments: {
selector: "select[name='country']",
value: "US"
}
});
kaboom_browser_upload — Upload file
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_upload",
arguments: {
selector: "input[type='file']",
filePath: "/path/to/file.pdf"
}
});
Accessibility Auditing
kaboom_a11y_audit — Run WCAG accessibility audit
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_a11y_audit",
arguments: {
standard: "WCAG2AA", // "WCAG2A", "WCAG2AA", "WCAG2AAA"
includeWarnings: true
}
});
kaboom_a11y_export — Export audit results
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_a11y_export",
arguments: {
format: "sarif" // "sarif", "json", "html"
}
});
Security Auditing
kaboom_security_audit — Run security audit
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_security_audit",
arguments: {
checkCredentials: true,
checkPII: true,
checkHeaders: true,
checkCookies: true,
checkThirdParty: true
}
});
Web Vitals Monitoring
kaboom_vitals_get — Get Web Vitals metrics
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_vitals_get",
arguments: {
metrics: ["LCP", "CLS", "INP", "FCP"] // Optional, defaults to all
}
});
kaboom_vitals_baseline — Set performance baseline
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_vitals_baseline",
arguments: {
metric: "LCP",
threshold: 2500 // milliseconds
}
});
Visual Annotations
kaboom_annotation_enable — Enable drawing mode
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_annotation_enable"
});
kaboom_annotation_get — Get annotations with computed styles
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_annotation_get",
arguments: {
includeStyles: true
}
});
Common Workflows
Debug Console Errors
// 1. Get all error logs
const errors = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_console_get",
arguments: { level: "error" }
});
// 2. Inspect DOM at error location
const elements = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_dom_query",
arguments: {
selector: ".error-component",
includeStyles: true
}
});
// 3. Clear logs after fix
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_console_clear"
});
Debug Failed API Calls
// 1. Get failed network requests
const failed = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_network_get",
arguments: {
status: "failed",
includeBody: true
}
});
// 2. Check for auth issues in request headers
// Analyze the response from failed requests
// 3. Clear network buffer
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_network_clear"
});
Record User Flow and Generate Test
// 1. Start recording
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_recording_start",
arguments: {
captureClicks: true,
captureInputs: true,
captureNavigation: true
}
});
// 2. User performs actions in browser...
// 3. Stop recording and get actions
const actions = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_recording_stop"
});
// 4. Generate Playwright test
const test = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_recording_generate_test",
arguments: {
format: "playwright",
includeAssertions: true
}
});
Automate Browser Testing
// Navigate to login page
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_navigate",
arguments: {
url: "https://app.example.com/login",
waitForLoad: true
}
});
// Fill in credentials
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_type",
arguments: {
selector: "input[name='email']",
text: "test@example.com"
}
});
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_type",
arguments: {
selector: "input[name='password']",
text: "testpass123"
}
});
// Submit form
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_click",
arguments: {
selector: "button[type='submit']"
}
});
// Check for errors
const errors = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_console_get",
arguments: { level: "error" }
});
Accessibility Audit
// Run WCAG 2.1 AA audit
const audit = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_a11y_audit",
arguments: {
standard: "WCAG2AA",
includeWarnings: true
}
});
// Export results as SARIF
const report = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_a11y_export",
arguments: {
format: "sarif"
}
});
Performance Regression Detection
// Set baseline for LCP
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_vitals_baseline",
arguments: {
metric: "LCP",
threshold: 2500
}
});
// Get current metrics
const vitals = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_vitals_get"
});
// Check if LCP regressed
if (vitals.LCP > 2500) {
console.log("LCP regression detected!");
}
Browser Extension Developer API
Use window.__kaboom in your application code for custom context:
// Add custom annotation to help AI understand context
window.__kaboom.annotate({
type: "error-boundary",
component: "CheckoutForm",
message: "Payment validation failed",
metadata: {
step: 3,
validationErrors: ["invalid_card", "expired"]
}
});
// Check if Kaboom is available
if (window.__kaboom) {
// Kaboom is active
}
Configuration
Environment Variables
# Disable telemetry
export KABOOM_TELEMETRY=off
# Custom port (default: 3333)
export KABOOM_PORT=8080
# Log level
export KABOOM_LOG_LEVEL=debug # debug, info, warn, error
# Custom extension path
export KABOOM_EXTENSION_PATH=/custom/path/to/extension
Server CLI Flags
# Start MCP server
kaboom --mcp
# Custom port
kaboom --mcp --port 8080
# Enable debug logging
kaboom --mcp --debug
# Print version
kaboom --version
Troubleshooting
Extension Not Connecting
-
Check extension is loaded:
- Open
chrome://extensions - Verify Kaboom is enabled
- Check for errors in extension details
- Open
-
Verify server is running:
ps aux | grep kaboom -
Check connection:
- Open browser DevTools
- Look for "Kaboom connected" in console
- If not connected, refresh the page
No Console Logs Captured
-
Clear cache and refresh:
await use_mcp_tool({ server_name: "kaboom", tool_name: "kaboom_console_clear" }); -
Check log level filter — Make sure you're not filtering out the logs you need
-
Verify page is active — Kaboom only captures from the active tab
Network Requests Not Appearing
-
Check status filter:
// Get ALL requests, not just failed await use_mcp_tool({ server_name: "kaboom", tool_name: "kaboom_network_get", arguments: { status: "all" } }); -
Verify XHR/Fetch interception — Some requests may be blocked by CSP
-
Check response body size — Very large bodies may be truncated
Recording Generates Weak Selectors
-
Add explicit IDs/data attributes to critical elements:
<button data-testid="submit-btn">Submit</button> -
Use semantic HTML —
<button>instead of<div onclick> -
Manually improve selectors in generated tests after recording
MCP Client Not Finding Kaboom
-
Check binary path:
which kaboom # Should output: /usr/local/bin/kaboom -
Update config with full path:
{ "mcpServers": { "kaboom": { "command": "/usr/local/bin/kaboom", "args": ["--mcp"] } } } -
Restart MCP client completely (not just reload window)
Performance Impact
Kaboom is designed for minimal overhead, but:
- Disable when not debugging — Unload extension if not needed
- Clear buffers regularly — Large console/network buffers can grow
- Use selective recording — Only record what you need
Privacy & Security
- All data stays local — Nothing leaves your machine
- No cloud, no accounts — Zero external dependencies
- Auth headers stripped — Sensitive headers automatically removed
- Localhost only — Server binds to
127.0.0.1only - Anonymous telemetry — Random install ID, no personal data (disable with
KABOOM_TELEMETRY=off)
Advanced Usage
Custom Test Generation Template
Generate tests with custom assertions:
const test = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_recording_generate_test",
arguments: {
format: "playwright",
includeAssertions: true
}
});
// Add custom assertions to generated test
const customTest = test.replace(
"// end of test",
`
// Custom assertions
await expect(page.locator('.success-message')).toBeVisible();
await expect(page).toHaveURL(/.*dashboard/);
`
);
Continuous Monitoring Pattern
// Set up baseline thresholds
const thresholds = {
LCP: 2500,
CLS: 0.1,
INP: 200,
FCP: 1800
};
for (const [metric, threshold] of Object.entries(thresholds)) {
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_vitals_baseline",
arguments: { metric, threshold }
});
}
// Periodically check for regressions
const vitals = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_vitals_get"
});
// Flag regressions
const regressions = Object.entries(vitals)
.filter(([metric, value]) => value > thresholds[metric])
.map(([metric, value]) => `${metric}: ${value} > ${thresholds[metric]}`);
if (regressions.length > 0) {
console.log("Performance regressions detected:", regressions);
}
Multi-Page Recording
// Start recording
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_recording_start",
arguments: { captureNavigation: true }
});
// Navigate through multiple pages
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_navigate",
arguments: { url: "https://example.com/page1" }
});
// Perform actions...
await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_browser_navigate",
arguments: { url: "https://example.com/page2" }
});
// More actions...
// Stop and generate comprehensive test
const recording = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_recording_stop"
});
const test = await use_mcp_tool({
server_name: "kaboom",
tool_name: "kaboom_recording_generate_test",
arguments: {
format: "playwright",
includeAssertions: true
}
});
Resources
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/reason-machines/devtools-skills/kaboom-browser-ai-devtools">View kaboom-browser-ai-devtools on skillZs</a>