check-proxy-tax

Find out what a "cheap" proxy actually costs per 1,000 usable requests.

Proxy invoices are priced per request, per IP, or per GB. Your scraper is not paid in those units. It is paid in responses it can actually use. When an edge returns 403, a request stalls until timeout, or a 200 arrives without the product marker, the attempt is still billed. The gap between the advertised price and the price of usable traffic is The 403 Tax.

check-proxy-tax measures that gap directly. It fires a small concurrent burst at one URL, classifies every response as usable or blocked, and reports:

• usable rate and blocked/failed counts;
• effective cost per 1,000 usable requests;
• the tax multiplier over the nominal per-1,000 price;
• p50 / p95 latency and total bytes transferred;
• status-code and failure-reason breakdown.

It is a single core module using only the standard library. No account, no API key, no external dependencies, no background service.

Install

Python 3.8 or newer.

git clone https://github.com/600me/check-proxy-tax.git
cd check-proxy-tax
python3 -m pip install .
check-proxy-tax --help

You can also run it straight from a checkout without installing:

python3 -m check_proxy_tax.cli --url https://example.com --requests 5

The package is PyPI-ready (python3 -m build produces a wheel and sdist), but installing from source is the supported path until the first PyPI release.

Quick start

check-proxy-tax \
  --url "https://example.com/product/123" \
  --requests 20 \
  --marker "Add to cart" \
  --price-per-1k 0.49

check-proxy-tax 0.1.0 - proxy economics audit

+---------------------+------------------------------------------+
| METRIC              | VALUE                                    |
+---------------------+------------------------------------------+
| Target URL          | https://example.com/product/123          |
| Samples             | 20 (concurrency 4, timeout 10s)          |
| Marker              | Add to cart                              |
| Usable success      | 14 / 20                                  |
| Blocked / failed    | 6 / 20                                   |
| Usable rate         | 70.0%                                    |
+---------------------+------------------------------------------+
| Nominal price / 1k  | $0.4900                                  |
| Effective cost / 1k | $0.7000                                  |
| Tax multiplier      | 1.43x                                    |
+---------------------+------------------------------------------+
| Latency p50         | 212.4 ms                                 |
| Latency p95         | 1840.9 ms                                |
| Total transferred   | 1.18 MB (1,234,567 bytes)                |
| Status codes        | 200:14, 403:5, timeout:1                 |
| Blocked reasons     | http_403:5, timeout:1                    |
| Verdict             | DEGRADED - 6 of 20 requests blocked      |
+---------------------+------------------------------------------+

Benchmark & provider matrix: https://scrape.mybskm.top/?utm_source=cli&utm_content=check-proxy-tax
FTC Disclosure: Benchmark links may include affiliate tracking.

What counts as usable

A sample is usable only when both conditions hold:

1. the final HTTP status is exactly 200;
2. when --marker is supplied, the response body contains that string (case-sensitive, matched against the raw UTF-8 bytes).

Everything else is blocked: 403, 429, other 4xx, 5xx, timeout, connection/TLS failure, or a 200 without the marker. Redirects are followed and the final status is classified.

That definition is deliberately strict. A 200 challenge page, a consent screen, or an empty shell is not inventory your scraper can use.

The math

usable_pct            = usable_count / total_requests * 100
effective_cost_per_1k = nominal_price_per_1k / (usable_pct / 100)
tax_multiplier        = effective_cost_per_1k / nominal_price_per_1k

If usable_count is 0, the effective cost is Infinite: no amount of nominal budget buys a usable response at that sample rate. The JSON output sets effective_cost_per_1k and tax_multiplier to null and cost_status to "infinite", which keeps the document valid JSON.

Worked example

You pay $0.49 per 1,000 requests. You sample 20 requests and 14 come back usable with the expected marker.

usable_pct            = 14 / 20 * 100 = 70.0%
effective_cost_per_1k = 0.49 / 0.70   = $0.70
tax_multiplier        = 0.70 / 0.49   = 1.43x

At 1,000,000 requests per month the invoice is $490. Because only 70% of attempts are usable, you need about 1,428,571 attempts to obtain 1,000,000 usable responses, which costs about $700. The 403 Tax is $210 per month on that volume.

The model assumes blocked attempts are billed, which is how per-request and per-IP plans normally work. If your provider bills only successful requests, set --price-per-1k to the effective rate you want to test and read the multiplier as a quality signal rather than a surcharge.

Options

Option	Default	Description
--url URL	required	Absolute http(s) URL to probe.
--requests N	20	Number of samples.
--marker STRING	none	String that must appear in the body for a sample to count as usable.
--price-per-1k FLOAT	0.49	Nominal price per 1,000 requests, in USD.
--proxy URL	none	HTTP(S) proxy, e.g. http://user:pass@host:port.
--concurrency N	4	Maximum in-flight requests.
--timeout SECONDS	10	Per-request timeout.
--json	off	Emit one JSON object and nothing else.
--version		Print the version and exit.
--help		Show help, including the FTC disclosure.

Exit codes

Code	Meaning
0	Every sampled request was usable.
1	Degraded: at least one request was blocked, at least one succeeded.
2	Failed: no request was usable.
3	Invalid arguments or URL.

This makes the CLI useful as a gate without parsing text: 0 is the only fully healthy result. A CI job that must not fail on a degraded sample can capture the code and inspect the JSON instead.

JSON output

--json writes a single JSON object to stdout, suitable for jq, dashboards, or a CI artifact. Diagnostics never share that stream.

{
  "tool": "check-proxy-tax",
  "version": "0.1.0",
  "url": "https://example.com/product/123",
  "marker": "Add to cart",
  "requests": 20,
  "concurrency": 4,
  "timeout_seconds": 10.0,
  "proxy_configured": false,
  "nominal_price_per_1k": 0.49,
  "usable_count": 14,
  "blocked_count": 6,
  "usable_pct": 70.0,
  "effective_cost_per_1k": 0.7,
  "tax_multiplier": 1.428571,
  "cost_status": "finite",
  "latency_ms": {
    "p50": 212.4,
    "p95": 1840.9,
    "min": 141.2,
    "max": 2190.7
  },
  "total_bytes": 1234567,
  "status_counts": {
    "200": 14,
    "403": 5,
    "timeout": 1
  },
  "blocked_reasons": {
    "http_403": 5,
    "timeout": 1
  },
  "exit_code": 1,
  "benchmark_url": "https://scrape.mybskm.top/?utm_source=cli&utm_content=check-proxy-tax",
  "ftc_disclosure": "FTC Disclosure: Benchmark links may include affiliate tracking."
}

status_counts uses the numeric HTTP status as a string, or timeout / error when no status was received. blocked_reasons uses http_<status>, marker_missing, timeout, or connection_error.

CI example

name: proxy-cost-gate
on:
  workflow_dispatch:
  schedule:
    - cron: "17 6 * * *"

jobs:
  audit:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: "3.12"
      - run: python3 -m pip install .
      - name: Audit effective proxy cost
        env:
          TARGET_URL: https://example.com/product/123
        run: |
          set +e
          check-proxy-tax \
            --url "$TARGET_URL" \
            --requests 30 \
            --marker "Add to cart" \
            --price-per-1k 0.49 \
            --json > proxy-report.json
          code=$?
          set -e
          cat proxy-report.json
          # Gate on usable rate, not on the raw exit code.
          jq -e '.usable_pct >= 95 and .cost_status == "finite"' proxy-report.json

Run the suite locally with:

python3 -m unittest test_cli.py -v

The tests use a local http.server; they do not touch the network.

Benchmark and provider matrix

Public benchmark and provider comparison:

https://scrape.mybskm.top/?utm_source=cli&utm_content=check-proxy-tax

FTC Disclosure: Benchmark links may include affiliate tracking.

License

MIT. See LICENSE.

Copyright (c) 2026 Kais / 600me.