TokenAPI Scan

AI API Relay Verification Tools Comparison

The AI API relay market is a mixed bag. Several "relay detection" tools have emerged. We compare them objectively across 8 dimensions to help you choose the right tool.

Data collected on 2026-07-30 · Based on publicly available features, anonymized · Feedback welcome

📊 Core Comparison Matrix

Dimension 🪞 TokenAPI Scan
tokenscanai.com		 Tool A
Broad coverage, UGC-driven		 Tool B
In-depth, curated		 Tool C
Sub-tool matrix
Tracked sites
Detection rigor
Data transparency
Pricing data
Official source baseline
Claude protocol coverage
OpenAI protocol coverage
Gemini protocol coverage

🎯 Different Tools for Different Needs

Each tool has its own positioning — there's no "absolute best," only "best for whom":

  • 🪞 TokenAPI Scan (this site): Curated + transparent + exclusive pricing data. For users who seriously want to buy a relay and need a trustworthy reference. Leaderboard threshold: ≥3 tests, each with confidence labels.
  • Tool A (broad coverage): UGC-driven, hundreds of entries. For users who want to check if a long-tail domain has been tested. Wide coverage but noisy — ~half of sites tested only once.
  • Tool B (in-depth): Rich detail pages, comprehensive price tables. For users who want an "encyclopedia-style" reference. Slower updates but high depth.
  • Tool C (sub-tool matrix): Collection of sub-domain tools. For developers who want cross-validation via multiple tools.
💡 Tip: Different tools have different strengths — combining them gives a more complete picture. Our strength is curated data credibility + real-time pricing transparency, making us a "trustworthy reference" in your decision process.

🤔 FAQ

Why does TokenAPI Scan only list 30+ providers, fewer than others?

Different strategy. Broad-coverage tools follow a "list everything + user voting" model, so sites tested only once can appear on the leaderboard. We follow a "curated + high-quality data" model — the main leaderboard only shows sites with ≥3 tests, each with confidence labels. Our candidate pool (180+ sites) is continuously probed; only those reaching a stability threshold are promoted to the curated list — it's slow but thorough.

What do confidence labels mean?

Under a Bayesian-weighted scoring system, a "95" from 2 tests and a "95" from 35 tests have vastly different credibility. Each detail page is clearly labeled: High (10+ tests) / Medium (5-9) / Medium-Low (3-4) / Low (1-2). This way, when comparing AI API relays, you can prioritize results with more samples and higher confidence.

Why are api.openai.com and api.anthropic.com included?

These are official source baselines (⭐ marked), used to calibrate our detection algorithm — official sources should score near 100, and if they don't, the algorithm itself is flawed. This is our openly labeled "gold standard" so you can verify whether the algorithm is trustworthy.

Is the pricing data really real-time?

We auto-scrape pricing from 6 major relay providers weekly, with a 1-hour local cache. We're one of the few tools that shows price snapshots + historical trends directly on detail pages. The scraping script is open-source on GitHub.