ProxyRank

Rank proxy providers with real-world benchmarks — not synthetic lab tests.

ProxyRank runs actual HTTP traffic through your proxy against your target URL, measures full network timing (DNS → TCP → TLS → TTFB → download), and helps you pick the fastest, most reliable provider for your workload. Credentials never leave your machine.

npx @santhoshse7en/proxy-rank

Opens the web UI at http://127.0.0.1:9480 and starts the local agent on http://127.0.0.1:9470.

Why ProxyRank?

Most proxy comparisons stop at marketing claims or a single latency number. ProxyRank is built for teams who need to answer harder questions:

• Which provider is most reliable on my target under real load?
• Where do failures happen — DNS, TLS, TTFB, timeouts, blocks?
• Can I share a diagnostic report with my proxy vendor so they can fix bad exit nodes?
• How do BrightData, Oxylabs, and others compare on the same target, same concurrency, same TPS?

Ranking weights stability first — success rate and tail latency (P95/P99) matter more than a lucky average.

How it works

┌─────────────┐     HTTP/SSE      ┌──────────────────┐     proxy      ┌─────────────┐
│  Browser UI │ ────────────────► │  Local Agent     │ ─────────────► │   Target    │
│  (React)    │ ◄──────────────── │  127.0.0.1:9470  │ ◄───────────── │   Website   │
└─────────────┘   live metrics    └────────┬─────────┘   responses    └─────────────┘
       │                                   │
       │ credentials (localhost only)      │ via your proxy
       └──────────────────────────────────►│

Browsers cannot route HTTP through proxies with full timing control. A local Node.js agent executes real requests and streams per-request metrics back to the UI over SSE.

Privacy: Credentials live in browser memory and are sent only to 127.0.0.1. Optional AES-GCM encryption for saved profiles in localStorage. Nothing is uploaded to ProxyRank servers.

Quick start

Run (recommended)

npx @santhoshse7en/proxy-rank

Or install globally:

npm install -g @santhoshse7en/proxy-rank
proxy-rank

Develop locally

git clone https://github.com/santhoshse7en/proxy-rank.git
cd proxy-rank
npm install
npm run dev

| Mode | Web UI | Agent | |------|--------|-------| | npm package | http://127.0.0.1:9480 | http://127.0.0.1:9470 | | Development | http://localhost:5173 | http://127.0.0.1:9470 |

Benchmark workflow

1. Provider — Pick from 15+ built-in providers or configure a custom proxy
2. Credentials — Provider-specific form; test connection before running
3. Target — URL, HTTP method, headers, query, body; import from cURL
4. Confirm — Request count, concurrency, timeout, target TPS, country validation
5. Run — Live dashboard: RPS, latency sparklines, per-request timeline, logs
6. Results — KPIs, charts, request table, failure diagnostics, insights
7. Compare — Rank runs side-by-side across providers
8. Export — Audience-specific reports (see below)

All metrics come from real execution. No mock or simulated data.

Features

Load testing controls

| Setting | What it does | |---------|----------------| | Request count | Total requests in the run | | Concurrency | Parallel in-flight requests per benchmark | | Target TPS | Pace request starts (achieved TPS shown separately) | | Timeout | Per-request limit | | Country / exit IP | Validate geo routing per provider |

One benchmark runs at a time. Parallelism is within a run (concurrency), not across multiple providers simultaneously.

Results dashboard

Per-run analysis includes:

• Success rate, min/avg/P50/P95/P99/max latency
• Network timing breakdown (DNS, TCP, TLS, TTFB, download)
• Per-request waterfall and timeline
• Exit IP, ASN, ISP, geo validation
• Failure taxonomy (auth, timeout, blocked, rate-limit, network, HTTP errors)
• Auto-generated insights and recommendations

Compare

Compare every completed run in history — filtered by target, provider, config, date, or individual run.

• Stability-weighted scoring — P95/P99 and success rate weighted heavily
• Latest per provider — fair head-to-head with one recent run each
• Mixed-config warning — alerts when runs used different load profiles
• Charts: radar, grouped metrics, scatter, box plot, latency trend, score bars
• Expandable diagnostics per run; CSV/JSON compare export

Each historical run is a separate row. Re-running the same provider creates a new entry identified by completion timestamp.

History & settings

• History — Search, filter, rerun, export any past benchmark
• Settings — Encrypted credential storage, default load profile, theme/accent, factory reset
• Command palette — ⌘K quick navigation

Exports

Each format serves a different audience:

| Export | Audience | Purpose | |--------|----------|---------| | PDF Report | Managers, procurement | Executive summary — downloadable PDF | | CSV (Raw Requests) | Developers, analysts | Every request for Excel filtering and charts | | JSON | Developers, automation | Full benchmark payload + metadata | | HAR | Developers | Request/response debugging | | Provider Diagnostic | Proxy vendor support/NOC | Root-cause package: error analysis, exit nodes, slowest requests, timeline, recommendations + JSON + CSV |

The Provider Diagnostic Report is designed to answer why performance degraded — not just what the success rate was.

Supported providers

Built-in profiles with credential forms and connection testing:

BrightData · Oxylabs · NetNut · Smartproxy · SOAX · ScraperAPI · ScrapeOps · Scrape.do · ScrapingBee · Crawlbase · Decodo · IPRoyal · Webshare · DataImpulse · Zyte · Digitap (internal) · Custom proxy

Provider definitions live in proxyrank-shared and can be extended.

Project structure

Monorepo (npm workspaces):

| Package | Path | Role | |---------|------|------| | @proxy-rank/frontend | frontend/ | React + Vite UI | | proxyrank-agent | agent/ | Fastify agent — benchmark execution, SSE | | proxyrank-shared | shared/ | Types, providers, validation | | @santhoshse7en/proxy-rank | packages/proxyrank-local/ | Published npm bundle (CLI + UI + agent) |

Routes

| Route | Page | |-------|------| | / | Benchmark wizard | | /results/:id | Benchmark results | | /compare | Provider comparison | | /history | Benchmark history | | /settings | Settings |

Scripts

| Command | Description | |---------|-------------| | npm run dev | Web UI + local agent (watch mode) | | npm run dev:web | Frontend only | | npm run dev:agent | Agent only | | npm run build | Production build (shared → agent → frontend) | | npm run build:package | Build publishable npm package | | npm run start:package | Run package CLI from repo | | npm run publish:package | Build and publish to npm | | npm run typecheck | Typecheck all workspaces |

Documentation

• npm package guide — build, publish, install, test
• Architecture — system design, data flow, storage model

Requirements

• Node.js 20+
• A proxy provider account (for real benchmarks)
• Network access from your machine to the proxy and target

License

MIT — see package license.