Skip to content
BugFence
Zero-Dependency TypeScript SDK · ~15 KB Lazy Core · Self-Hosted or Cloud

Visual bug reports with developer diagnostics. Redact before upload.

BugFence adds an in-app reporting studio for annotated screenshots, recordings, and optional rolling 30-second DOM replay. Sensitive fields can be masked in the browser before report data is sent to your BugFence server.

  • Zero extension wall: nothing for reporters to install
  • 100% Self-hostable in your private VPC
  • Browser DOMFence redaction before upload
  • MCP server for AI coding agents
Privacy-By-Construction Flow

How Diagnostic Evidence Travels Safely to Your VPC

Client-Side Sanitization • TLS 1.3 Transport
STEP 01

In-Browser DOM Fencing

Passwords, payment card numbers (validated via Luhn algorithm), and elements with data-bugfence-mask are obscured locally before canvas rasterization or replay serialization.

Zero unredacted pixels leave device
STEP 02

Direct VPC Ingestion

Sanitized reports travel directly to your self-hosted API. Ingestion re-scrubs incoming payloads, validates release metadata, and stores records in your private PostgreSQL database or encrypted volume.

100% data sovereignty • Zero central cloud
STEP 03

Triage & Agent Dispatch

Engineers review symbolicated traces and 30s session replays rendered in a script-less sandboxed iframe. AI coding agents inspect reports via the built-in MCP server to patch regressions fast.

MCP Server for Claude Code, Cursor & Copilot
Want to test this in action? Experience real defect triggers and client-side masking in our sandbox.
View Sandbox Guide & Scenarios
Interactive Sandbox

Explore Both Sides of BugFence Live

See how non-technical reporters capture visual defects from the web application, and how developers triage them in the engineering console. Open each environment in a new tab to see them work together in real time.

Reporter Experience · QA & UsersPublic demo

NovaCloud SaaS Demo Application

Test the floating BugFence reporter widget embedded in a realistic SaaS app. Annotate screenshots, record audio/video, and experience client-side Luhn credit card scrubbing before data leaves the browser.

  • Floating in-app widget (no browser extensions needed)
  • Automatic client-side DOM & credit card sanitization
  • Interactive failure triggers to simulate real production bugs
Triage Experience · Developers & LeadsShared guest access

Developer Triage Console

Open the dashboard as a shared, read-only guest. Explore seeded reports, issue grouping, stack traces, and available replay examples without creating an account.

  • One-click passwordless access to a read-only guest workspace
  • 30-second session replay player in sandboxed viewer
  • Sample reports are shared with other demo visitors

8 Enterprise Defect Simulation Scenarios

Click any scenario below to launch it directly in the Demo Application in a new window, or test them inside the app.

Network / PII

Checkout Gateway Timeout (504)

Dispatches $2,450 payment with credit card numbers and bearer tokens. Simulated 2500ms upstream settlement timeout.

Telemetry:Client scrubs credit card with Luhn algorithm and masks bearer tokens before telemetry dispatch.

Launch Scenario in Demo App
Performance

Thread Freeze & Long Task (650ms)

Synchronous compute loop blocks the main JavaScript thread, causing the in-page FPS meter to drop.

Telemetry:PerformanceObserver captures LongTask violation breadcrumbs and frames dropped in session replay.

Launch Scenario in Demo App
GraphQL

GraphQL Schema Validation (400)

Malformed mutation query submitted with missing orderId and invalid plan tier enum values.

Telemetry:SDK categorizes response as GraphQL ValidationError and captures mutation variables with scrubbed PII.

Launch Scenario in Demo App
Runtime Exception

Deep TypeError Component Crash

Throws an unhandled TypeError with a 5-level component call stack in simulated unmounted settings.

Telemetry:Symbolicated stack trace captured with component boundary context for automated issue grouping.

Launch Scenario in Demo App
Security / CORS

Partner CRM Cross-Origin Block (403)

Outbound sync dispatch to legacy customer endpoint rejected by strict browser CORS policy.

Telemetry:Network interceptor logs failed preflight OPTIONS request and CORS restriction code.

Launch Scenario in Demo App
User Friction

Rage Click Cluster & 30s Replay

Rapidly clicking "Export Full Audit Log" triggers UI shake animation and marks high-friction interaction.

Telemetry:4+ rapid clicks clustered within 1.2s recorded in 30-second DOM replay timeline.

Launch Scenario in Demo App
Visual Stability

Cumulative Layout Shift (CLS 0.28)

Dynamic banner insertion suddenly pushes main viewport contents down by 140px without warning.

Telemetry:PerformanceObserver captures LayoutShift entry with high CLS delta and affected DOM elements.

Launch Scenario in Demo App
Deployment Drift

Stale CDN ChunkLoadError (404)

Dynamically importing invoice module fails after edge CDN invalidates older chunk hashes.

Telemetry:Automated triage tags ChunkLoadError and correlates with active release version.

Launch Scenario in Demo App

Ephemeral demo storage

Demo report media uses a temporary store with a 50 MB media quota and a 30-report FIFO limit. A 15-minute cleanup sweep removes reports older than 45 minutes. This policy applies to the demo tenant; your deployment’s storage policy is configured separately.

Open Triage Console

Developer & QA Pain Points

Why Traditional Bug Reporting Fails

Common friction points engineering teams experience when collecting bug reports from real users and clients.

The extension wall

Customers and enterprise users often cannot install browser extensions. BugFence is a drop-in widget: nothing for the reporter to install.

Privacy worries

Screenshots and recordings can capture tokens, emails and card numbers. BugFence masks known patterns and fenced fields in the browser, before upload.

Duplicate floods

Many users hit the same bug. Reports are grouped into issues by stack trace, endpoint, trace id or text similarity, with an explanation for each match.

Missing context

A screenshot alone is rarely enough. Reports carry console, failed requests, breadcrumbs, release and correlation ids (Sentry, Datadog, PostHog and more).

End-to-End Workflow

From Bug Capture to Verified Fix

01

Reporter Captures

User opens the in-app studio to capture an annotated screenshot, screen recording, or 30s DOM replay with zero extension needed.

02

Browser Redacts

DOMFence masks password inputs, cards, and custom selectors before rasterization. Tokens and emails are scrubbed before upload.

03

Server Groups & Symbolicates

Your self-hosted server groups duplicate reports into single issues using stack traces and release source maps, detecting regressions.

04

Agent or Team Fixes

Dispatch the issue to GitHub, Jira, or Linear, or let your AI coding agent (Claude Code, Cursor) resolve it over MCP with scoped read keys.

Interactive Privacy Playground

Experience Client-Side Redaction in Real Time

Edit the raw data on the left or test sample checkout payloads. Watch how BugFence scrubs credentials directly in the browser before any network dispatch.

Try a Scenario:
Editable

Executed entirely on your device. Zero network requests.

Sanitized Diagnostic Payload Fenced
Failed payment transaction for customer [EMAIL]
Order ID: ord_89234
Card Number: [CARD] (Luhn valid)
Authorization: Bearer [TOKEN]
Phone contact: [PHONE]
Findings Detected:1 Bearer token1 Email1 Card number1 Phone

Honest Notice: Pattern detection is best-effort. Screen video recordings include in-studio region redaction and timeline trimming controls. Form inputs and DOM trees can be explicitly protected via data-bugfence-mask.

Built for AI Coding Agents

Model Context Protocol (MCP) & CLI Native

Bug reports shouldn't just sit in a web dashboard. Give Claude Code, Cursor, and Copilot direct, privacy-sanitized diagnostic access.

AI Triage Without Leaking Secrets

Most bug reporting tools require pasting customer screenshots or raw database dumps into LLMs. BugFence provides a dedicated Model Context Protocol (MCP) server that runs locally via stdio or authenticated HTTP with read-only scoped keys.

  • Coding agents inspect symbolicated stack traces and failed API calls
  • Sensitive tokens and credentials stripped before prompt construction
  • On-device heuristics for free; bring-your-own LLM key for server triage
Claude Code / Cursor terminal
mcp:read
> claude --use-mcp bugfence
User: Investigate checkout error reported on release v2.4.1
Agent: Querying BugFence MCP tool `get_issue_diagnostics`...
{
  "issueId": "iss_9a2f",
  "failingEndpoint": "POST /api/checkout (500)",
  "symbolicatedStack": "src/checkout.ts:84 in processPayment",
  "redactedState": "authToken: [TOKEN], card: [CARD]"
}
Agent: Identified null check bug in src/checkout.ts line 84. Generating patch...

Comprehensive Feature Set

Built for Deep Technical Triage, Not Just Capture

Reporting Studio

Screenshot, tab recording and DOM replay in one dialog, with annotation tools, pixelate/redact and an element picker.

Redaction before upload

Password and card inputs are masked by default. You add selectors; tokens, emails, cards and more are scrubbed from text, URLs and headers. A live Privacy check shows what will change.

Reporter consent

Reporters can include or exclude screenshot, video, replay, console and network data before sending.

Issue grouping & regressions

Explainable grouping, priority ranking, regression detection, merge and detach.

Releases & source maps

Tag reports with release and environment, upload source maps per release, and symbolicate stacks on your server.

Agent-ready

An MCP server and CLI let coding agents (Claude Code, Cursor, VS Code/Copilot) read reports with read-only, scoped keys by default.

Integrations

GitHub, Jira, Linear, Slack, Teams, Discord, Datadog, Splunk, PagerDuty and signed webhooks.

Self-hosted

Run it on your own infrastructure with JSON files or PostgreSQL. Reports stay where you deploy them.

Ecosystem & Workflow

Integrates Seamlessly With Your Stack

BugFence connects with your current ticketing, observability, and alerting tools without vendor lock-in.

Issue Tracking

  • GitHub & GitHub Enterprise
  • Linear
  • Jira Cloud & Data Center

Session Correlation

  • Sentry Event & Replay IDs
  • PostHog Session URLs
  • LogRocket & FullStory

Monitoring & Logs

  • Datadog RUM / APM IDs
  • Splunk HEC
  • W3C Traceparent Context

Alerts & Webhooks

  • Slack Notifications
  • Microsoft Teams
  • Discord & HMAC Webhooks

Developer Quickstart

Add BugFence to Your App in 5 Minutes

Use your public project key in the browser. Technical context and privacy redaction are active immediately.

// 1. Install
npm install @social-pulse-crm/bugfence-react

// 2. Wrap your application with BugFenceProvider
import { BugFenceProvider } from '@social-pulse-crm/bugfence-react';

export function RootLayout({ children }: { children: React.ReactNode }) {
  return (
    <BugFenceProvider
      config={{
        projectKey: 'pk_live_your_public_key',
        endpoint: 'https://bugs.yourcompany.com/api/v1/reports',
        lazyWidget: true,
        accentColor: '#0b7a5f',
        features: {
          screenRecording: true,
          voice: true,
        },
      }}
    >
      {children}
    </BugFenceProvider>
  );
}
Only public keys are exposed to client code. Secrets and LLM API keys stay safely on your server.
View @social-pulse-crm/bugfence-react on NPM

Frequently asked questions

How does redaction work for screenshots, DOM replays, and video recordings?

BugFence provides defense-in-depth privacy across all capture types: DOM screenshots and 30s session replays automatically mask passwords, credit cards, and tagged selectors (data-bugfence-mask). For screen video recordings, the studio includes built-in video redaction tools (drag-to-redact region boxes across custom timeframes), start/end timeline trimming, and device-side re-encoding, allowing reporters to review and redact any sensitive pixels before attachment.

Do reporters need an extension?

No. BugFence is a script or React component embedded in your app.

Where is data stored?

On infrastructure you run. The documented deployment model is self-hosted (Docker, Node, JSON files or PostgreSQL).

Does AI see my screenshots?

On-device heuristics make no network calls. Optional server-side AI triage sends only a minimal, re-sanitised text subset to the provider you configure; screenshots, video and replay are never sent.

Does it work on mobile?

The widget runs in mobile browsers, but screen recording is generally unavailable there because browsers do not expose getDisplayMedia.

Ready for Production SaaS

Stop Guessing Bugs. Start Debugging Real Evidence.

Integrate the zero-dependency widget in 5 minutes. No reporter extensions, browser-side masking controls, and a self-hosted deployment option.