Use Cases — ProceedGate

The Problem

Your coding assistant processes tickets using GPT-4 ($0.03/1K tokens). A buggy ticket causes infinite reasoning — the agent retries 15+ times. $45+ burned in 3 minutes.

With ProceedGate

Each LLM call goes through gate.check(). After 10 identical calls in 60s, loop detection triggers → 429 BLOCKED. Agent is stopped. $35+ saved.

// Gate every LLM call
const decision = await gate.check({
  action: 'openai.gpt4.completion',
  context: {
    model: 'gpt-4-turbo',
    estimated_tokens: 2500,
    estimated_cost_usd: 0.075,
  }
});

if (decision.status === 200) {
  await openai.chat.completions.create({...});
} else if (decision.status === 429) {
  // Storm detected — ProceedGate blocked the loop
  log.error('Agent loop detected, stopping');
}
        

Try it live — Hit the real API

Click "Test API" to call the live ProceedGate API. Click 10+ times rapidly to trigger storm detection!

The Problem

BNB drops 15% in 30 seconds. Your arb bot panic-executes 20 swaps at escalating gas prices, losing $200+ in gas + slippage.

With ProceedGate

When 10+ identical pair swaps fire in 60s, storm detection kicks in → 429 BLOCKED. Bot is paused. Gas + slippage saved.

// Gate every DEX swap
const decision = await gate.check({
  action: 'defi.pancakeswap.swap',
  context: {
    pair: 'BNB/USDT',
    amount_usd: 5000,
    gas_estimate_usd: 2.50,
    spread_percent: -8.0  // negative = losing trade
  }
});
// 429 → Bot stops and waits for market to stabilize
        

Try it live

Click "Test API" to simulate a DeFi swap check

The Problem

CryptoScraper hits a 403 error. Built-in retry logic hammers the API 50 times before backoff kicks in. $0.001/req × 50 req × 100 tasks = unexpected bill.

With ProceedGate

Detects >10 identical requests/minute. Storm blocked after call #10. Agent gets a 429 + reason, logs it, and moves on. 90% of wasted calls prevented.

// Gate every scraping request
const decision = await gate.check({
  action: 'scraper.coinmarketcap.fetch',
  context: {
    url: 'https://api.coinmarketcap.com/v2/...',
    estimated_cost_usd: 0.001,
    retry_count: 12  // ProceedGate sees this climbing
  }
});
        

Try it live

Click "Test API" to simulate a scraping check

The Problem

4-agent pipeline: Researcher → Analyst → Writer → Reviewer. Researcher enters a disambiguation loop, searching repeatedly. SerpAPI credits drain.

With ProceedGate

All 4 agents share one workspace = one budget = one storm detector. When Researcher loops, ProceedGate blocks it without affecting others. Pipeline continues.

// All agents share one workspace
const gate = createGate({ workspaceId: 'ws_research_pipeline' });

// Each agent identifies itself
await gate.check({
  action: 'serpapi.search',
  context: {
    agent: 'researcher',
    query: 'AI agent cost overruns 2025',
    pipeline: 'research_report'
  }
});
        

Try it live

Click "Test API" to simulate a multi-agent check

The Problem

A Zapier zap triggers on form submission → GPT-4 → Twilio SMS. A bot spams your form 500 times. 500 GPT-4 calls + 500 SMS = $200+ surprise.

With ProceedGate

Add a ProceedGate webhook step before expensive actions. Rate limiting + loop detection per source. Bot spam blocked after 10 identical triggers. $190+ saved.

// n8n webhook → ProceedGate → GPT-4 → Twilio
// HTTP Request node in n8n:
POST https://governor.proceedgate.dev/v1/workspaces/{{wsId}}/check
{
  "action": "automation.form_to_gpt4_sms",
  "context": {
    "source_ip": "{{$json.ip}}",
    "estimated_cost_usd": 0.04
  }
}
        

Feature Matrix — What ProceedGate Catches

Feature	LLM Gateway	DeFi Bot	Scraping	Multi-Agent	Webhooks
Loop detection (>10 identical/min)	✓	✓	✓	✓	✓
Per-workspace budget cap	✓	✓	✓	✓	✓
Proceed token (JWT)	✓	✓	✓	✓	✓
Webhook alerts	✓	✓	✓	✓	✓
Multi-agent shared budget	—	—	—	✓	✓