ProceedGate predicts retry loops 30 seconds early and blocks them before they drain your API credits. Start in Shadow Mode โ see what would be blocked, zero enforcement risk.
Every team running scraping or AI agents hits these โ usually at 3 AM.
One failed API call triggers 500 retries in under a minute. Loops burn budget in minutes.
No guardrails means wasted credits. Dashboards show the spend after it happened.
Monitoring tools show you the fire. ProceedGate stops it before it starts.
Not mockups. Real API response structures from production.
Agent: my-agent Pattern: 47 identical requests / 60s Zone: RED Action: BLOCKED Credits saved: 142 AI reasoning: baseline_deviation: +4.2ฯ behavior_entropy: 0.12 (low โ repetitive) reputation_tier: untrusted throttle_level: 5/5 (blocked) Prediction: trend: accelerating (slope +2.3 RPM/s) storm_eta: ~8 seconds confidence: 0.94 Webhook โ Slack #alerts โ
$ curl governor.proceedgate.dev/v1/check { "decision": "proceed", "zone": "green", "credits_remaining": 1847, "reputation": { "tier": "trusted" }, "prediction": { "direction": "stable", "confidence": 0.87 }, "throttle": { "level": 0, "label": "none" }, "latency_ms": 12 }
npm install @proceedgate/sdk
Add governor.gate() before expensive operations
Per-agent limits, per-task caps โ AI handles the rest
import { CostGovernor } from '@proceedgate/sdk';
const gov = new CostGovernor({ apiKey: process.env.PG_KEY });
const { decision } = await gov.gate({
agent: 'scraper-v2',
action: 'api_call',
cost: 0.03,
});
if (decision === 'proceed') {
await callExpensiveAPI();
}One command. 30 seconds. See what ProceedGate would block in your workflow.
## Install and run a quick check $ npx @proceedgate/runner check --workspace demo Scanning agent: scraper-v2... Requests analyzed: 147 Loops detected: 3 Estimated savings: $42.80 Shadow blocks: 12 (would have blocked) โ Report saved to ./proceedgate-report.json
Not rules. Not heuristics. Real ML models running on every request โ explainable, auditable, open source.
EWMA + z-score + CUSUM. Learns normal, detects anomalies.
Shannon entropy, burst index, sequence hashing.
Storm warnings 30s early via trend regression.
6-level graduated response. Not binary allow/block.
Llama 3.1 8B for ambiguous cases. Full reasoning chains.
5-component trust. Good agents get lenient thresholds.
Global anomaly detection. Differential privacy (ฮต=1.0).
Inspect every model, score, prediction. Full transparency.
Apify, SerpAPI, Firecrawl, Bright Data โ retry storms are the #1 cost killer.
Deep dive โLangChain, AutoGPT, CrewAI โ any pipeline calling external APIs can silently loop.
Background AI workflows โ email, content, data processing โ that can fail-loop silently.
Per-client budgets, per-agent limits. Hundreds of agents, zero rogue tasks.
Trusted by dev teams building:
Start in Shadow Mode. See what ProceedGate would block โ without touching your running agents. Switch to enforcement when you trust the data.
const gov = new CostGovernor({
apiKey: process.env.PG_KEY,
mode: 'LOG_ONLY', // shadow mode
});
const result = await gov.gate({
agent: 'scraper-v2',
action: 'api_call',
cost: 0.03,
});
// Always proceeds, but logs
// what it would block
if (result.would_block) {
console.warn('Storm detected');
}Direct access to founder
Influence the roadmap
No lock-in. Cancel anytime.
Can't find your answer? Reach out โ we reply fast.