kaboom

name	kaboom
description	Use when user asks to check browser state, debug page errors, inspect network traffic, take screenshots, automate clicks or form fills, run accessibility or security audits, measure performance, generate tests, or record browser sessions. Trigger phrases: "check my browser", "debug this page", "take a screenshot", "browser errors", "network requests", "automate the browser", "record a test", "accessibility audit", "performance check", "what's on the page".
license	AGPL-3.0
compatibility	Requires kaboom-agentic-browser binary on PATH and Kaboom Chrome extension connected. macOS, Linux, Windows.
metadata	{"author":"Kaboom AI","version":"0.8.4","category":"developer-tools","tags":["browser","debugging","automation","testing","observability"]}

Kaboom

Critical: Daemon Setup

Before any tool call, ensure the daemon is running:

bash scripts/ensure-daemon.sh

If this fails, run the connection doctor:

bash scripts/connection-doctor.sh

How to Call Tools

All 5 tools are called via the helper script:

bash scripts/kaboom-call.sh <tool> '<json_arguments>'

Examples:

# Read browser errors
bash scripts/kaboom-call.sh observe '{"what":"errors","limit":20}'

# Take a screenshot
bash scripts/kaboom-call.sh observe '{"what":"screenshot","format":"png"}'

# Click an element
bash scripts/kaboom-call.sh interact '{"what":"click","selector":"button.submit"}'

# Run accessibility audit
bash scripts/kaboom-call.sh analyze '{"what":"accessibility"}'

# Generate a Playwright test
bash scripts/kaboom-call.sh generate '{"what":"test","test_name":"login_flow"}'

# Check health
bash scripts/kaboom-call.sh configure '{"what":"health"}'

Tools Quick Reference

observe — Read captured browser telemetry (passive)

Top modes: errors, logs, network_waterfall, network_bodies, actions, screenshot, page, tabs, storage, recordings, websocket_events, error_bundles, timeline, transients, page_inventory

Common params: what (required), limit, since_cursor, summary, format, url, scope

Pagination: pass after_cursor/before_cursor/since_cursor from response metadata. Set restart_on_eviction=true if cursor expired.

Full reference: references/observe-modes.md

analyze — Active analysis queries (sent to extension)

Top modes: dom, accessibility, security_audit, performance, forms, link_health, page_summary, computed_styles, visual_diff, api_validation, third_party_audit, audit

Common params: what (required), selector, timeout_ms, background, summary

Async mode: set background:true to get correlation_id, then poll with observe(what='command_result', correlation_id=...).

Full reference: references/analyze-modes.md

generate — Create artifacts from captured data

Top modes: test, reproduction, pr_summary, har, csp, sarif, visual_test, test_from_context, test_heal, test_classify

Common params: what (required), test_name, save_to, last_n

Full reference: references/generate-modes.md

configure — Session management and diagnostics

Top modes: health, doctor, store, clear, describe_capabilities, event_recording_start, event_recording_stop, playback, noise_rule, streaming, telemetry, diff_sessions

Common params: what (required), store_action, key, value

Full reference: references/configure-modes.md

interact — Browser automation

Top modes: click, type, navigate, screenshot, execute_js, explore_page, list_interactive, fill_form, scroll_to, hover, wait_for, batch

Common params: what (required), selector, text, url, script

Targeting priority: selector > element_id > index > x/y coordinates

Semantic selectors: text=Submit, role=button, placeholder=Email

Composable enrichment: add include_screenshot:true, action_diff:true, or wait_for_stable:true to any action.

Full reference: references/interact-actions.md

Workflows

Choose the workflow that matches your task:

Task	Workflow
Debug errors, triage failures, create regression tests	references/workflow-debug-and-triage.md
Site audit, UX review, accessibility, security, API validation	references/workflow-audit-and-security.md
Browser automation, demos, test generation, release gates	references/workflow-automate-and-test.md
Performance measurement and comparison	references/workflow-performance.md

Troubleshooting

Daemon not running

Run: bash scripts/ensure-daemon.sh If fails: check which kaboom — binary must be on PATH. Install: npm install -g kaboom-agentic-browser

Extension not connected

Run: bash scripts/connection-doctor.sh Check Chrome extension is installed and enabled on the target page. The extension POSTs telemetry to localhost:7890.

No data returned

Ensure a tab is tracked:

bash scripts/kaboom-call.sh observe '{"what":"tabs"}'

If no tracked tab, open a page with the extension active.

Tool call returns error

Use describe_capabilities for parameter help:

bash scripts/kaboom-call.sh configure '{"what":"describe_capabilities","tool":"observe","mode":"errors"}'

Connection lost mid-session

Re-run ensure-daemon and retry:

bash scripts/ensure-daemon.sh
bash scripts/kaboom-call.sh configure '{"what":"health"}'