Glitchgrab

Grab the glitch. Ship the fix.

Glitchgrab turns messy bug reports — screenshots, production errors, user complaints — into well-structured GitHub issues using AI.

npm: npm install glitchgrab | Website: glitchgrab.dev | SDK Docs: packages/sdk-nextjs/README.md

What Makes Glitchgrab Different?

There are existing tools in this space (see Competitive Landscape), but Glitchgrab combines unique capabilities that no single tool offers today:

Capability	Jam.dev	Marker.io	BetterBugs	Sentry	Glitchgrab
Screenshot → GitHub issue	✅	✅	✅	❌	✅
SDK auto-captures production errors	❌	❌	Partial	✅	✅
End-user "Report Error" button	❌	Widget	Widget	❌	✅
AI-generated issue (title, body, labels)	❌	❌	AI assist	❌	✅
Dedup check before creating issue	❌	❌	❌	✅	✅
MCP server (Claude integration)	❌	❌	❌	❌	✅
Open source	❌	❌	❌	✅	✅

Key differentiators

1. AI-first issue generation: Not just AI-assisted — the AI writes the entire issue (title, description, labels, severity) from raw input.
2. Smart dedup & merge: AI compares new reports against open issues. Similar bugs get added as comments, not new issues.
3. MCP server: Query and create issues from Claude directly. No other bug tool has this.
4. Open source: Full codebase available. Built by developers, for developers.

Three Components

1. SDK (glitchgrab)

// app/layout.tsx
import { GlitchgrabProvider } from "glitchgrab";

export default function RootLayout({ children }) {
  return (
    <GlitchgrabProvider token="gg_your_token">{children}</GlitchgrabProvider>
  );
}

• Auto-captures unhandled errors with full context (visited pages, stack, screenshot)
• Adds a "Report Error" button for end-users
• Non-blocking — never crashes the host app

2. Web Dashboard

• Connect GitHub repos, generate tokens (one token = one repo)
• Upload screenshots → AI creates issues
• Chat-based issue creation — describe a bug, AI handles the rest
• View all reported issues across repos
• Available at glitchgrab.dev

3. MCP Server

• Connect to Claude Desktop or any MCP client
• "What bugs were reported on my-app today?"
• "Create a feature request for dark mode"

Competitive Landscape

Tool	What it does	Pricing	Limitations
Jam.dev	Chrome extension for instant bug reporting with screenshots, console logs, network data	Free + paid tiers	Browser extension only, no SDK, no AI issue generation
Marker.io	Website feedback widget with annotations, metadata capture	From $39/mo	Focused on visual feedback/QA, no error auto-capture SDK, no AI generation
BetterBugs	Chrome extension + Web SDK for screenshot/screen recording with AI debugging assistant	Free tier + paid	Has SDK but focused on manual reporting, AI is for debugging not generation
Sentry	Full observability platform — error tracking, performance, session replay	Free tier + paid	Heavyweight, no AI issue creation, doesn't create GitHub issues auto

Where Glitchgrab fits

Glitchgrab is not trying to be Sentry (full observability) or Marker.io (agency feedback workflows). It's a focused tool that solves one problem: the gap between noticing a bug and having a well-written GitHub issue.

Tech Stack

Layer	Technology
Framework	Next.js 15 (App Router)
Database	Neon (Serverless PostgreSQL)
ORM	Prisma
Auth	NextAuth.js (GitHub OAuth)
AI	Claude API + OpenAI API
Deployment	Vercel
Monorepo	Turborepo
Package Manager	bun

Project Structure

glitchgrab/
├── apps/
│   ├── web/                    # Next.js dashboard + API
│   └── mobile/                 # React Native (Expo) mobile app
├── packages/
│   ├── sdk-nextjs/             # glitchgrab npm package
│   ├── mcp-server/             # MCP server for Claude
│   └── shared/                 # Shared types
├── CLAUDE.md                   # Instructions for Claude Code
├── README.md
├── package.json
└── turbo.json

Getting Started

git clone https://github.com/webnaresh/glitchgrab.git
cd glitchgrab
bun install
cp apps/web/.env.example apps/web/.env.local
# Fill in your keys
bun run dev

Roadmap

• GitHub OAuth + repo connection + token generation
• AI pipeline: image/text → structured GitHub issue
• SDK: Error boundary + auto-capture
• SDK: Report Error button
• Dashboard: Upload screenshots
• AI deduplication check
• MCP server
• Mobile app (Android + iOS)
• Linear + Jira support
• Multi-framework SDK (React, Vue)

License

MIT

---

Built by Navibyte Innovation Pvt. Ltd.