Glass Box - AI Agent Observability Platform

See inside your AI agents. Track tokens, costs, latency, and debug failures with a developer-first observability platform.

Features

• Real-time Logging - Stream events from your agents with our simple API
• Cost Tracking - Monitor token usage and costs across all your agents
• Performance Metrics - Track latency, cache hits, and success rates
• Timeline View - Debug agent runs step-by-step
• Multi-tier Billing - Flexible plans from hobbyists to enterprise

Quick Start

1. Clone and Install

git clone https://github.com/yourusername/glass-box.git
cd glass-box
npm install

2. Set up Supabase

1. Create a new Supabase project at supabase.com
2. Run the migration in supabase/migrations/001_initial.sql in the SQL editor
3. Copy your project URL and keys to .env.local

3. Set up Stripe (Test Mode)

1. Create a Stripe account and get test keys from the dashboard
2. Create 3 products with monthly prices:
   • Hacker: $29/mo
   • Startup: $99/mo
   • Scale: $299/mo
3. Add price IDs to .env.local
4. Set up webhook endpoint: https://yoursite.com/api/stripe/webhook
5. Listen for events: checkout.session.completed, customer.subscription.*

4. Configure Environment

cp .env.example .env.local
# Edit .env.local with your keys

5. Seed Demo Data (Optional)

npm run seed

6. Run Development Server

npm run dev
# Visit http://localhost:3000

API Usage

Authentication

Get your API key from the Settings page after signing up.

Ingest Events

curl -X POST https://app.glassbox.dev/api/ingest \
  -H "Content-Type: application/json" \
  -H "x-glassbox-key: YOUR_API_KEY" \
  -d '{
    "request_id": "req_123",
    "step_type": "llm",
    "tool_name": "gpt-4",
    "input_tokens": 150,
    "output_tokens": 200,
    "cost_usd": 0.0105,
    "latency_ms": 1200,
    "cache_hit": false,
    "status": "ok",
    "payload": {
      "model": "gpt-4",
      "temperature": 0.7
    },
    "index": 0
  }'

Node.js SDK Example

import { GlassBox } from 'glassbox-sdk'; // Coming soon

const gb = new GlassBox({ apiKey: process.env.GLASSBOX_KEY });

// Log a single step
await gb.log({
  requestId: 'req_123',
  stepType: 'llm',
  toolName: 'gpt-4',
  inputTokens: 150,
  outputTokens: 200,
  costUsd: 0.0105,
  latencyMs: 1200,
  status: 'ok'
});

// Or use the tracer for automatic tracking
const tracer = gb.trace('req_123');
const response = await tracer.llm('gpt-4', { 
  messages: [...] 
});

Deployment

Deploy to Vercel

1. Fork this repository
2. Import to Vercel
3. Add environment variables
4. Deploy!

Production Checklist

• Enable Supabase RLS policies
• Switch Stripe to live mode
• Set up monitoring (Sentry, PostHog)
• Configure rate limiting
• Set up backup strategy
• Add custom domain
• Enable SSL certificate

Architecture

┌─────────────┐     ┌──────────────┐     ┌─────────────┐
│   Next.js   │────▶│   Supabase   │────▶│  Postgres   │
│  App Router │     │   Auth/RLS   │     │   Database  │
└─────────────┘     └──────────────┘     └─────────────┘
       │                                         │
       ▼                                         ▼
┌─────────────┐                         ┌─────────────┐
│    Stripe   │                         │   Events    │
│   Billing   │                         │    Table    │
└─────────────┘                         └─────────────┘

Testing

Test Cards (Stripe)

• Success: 4242 4242 4242 4242
• Decline: 4000 0000 0000 0002
• 3DS Required: 4000 0025 0000 3155

Rate Limits

• Ingest API: 60 requests/minute per key
• Dashboard: No limits
• Metrics: Cached for 60 seconds

Support

• Documentation: docs.glassbox.dev
• Discord: discord.gg/glassbox
• Email: support@glassbox.dev

License

MIT

---

Built with ❤️ for the AI agent community.