How we test

Short version: we read the vendor's own pages, we write down exactly what they say, we date it, and we never let a sponsor move a ranking.

Current methodology (launch edition, October 2026)

At launch, every comparison on BestAgentStack is researched from official documentation: the vendor's published pricing page, product pages, and developer docs. For each tool we record the same fields:

• Starting paid price — the lowest-cost way to pay for the product, quoted exactly as the vendor states it, including what volume or usage it covers.
• Free tier — whether one exists and its concrete limits (emails per month, free transactions, free credits). A time-limited trial is labeled a trial, not a free tier.
• One notable strength or limit — the single fact most likely to change an agent builder's decision: a pricing-model quirk, an inbound-handling gap, a spending-control model, a missing pricing page.
• Checked date — the month and year the facts were read on the official page. Prices change; the date tells you how much to trust the number today.

Hands-on API tests — signing up, sending real requests, measuring behavior — are rolling out category by category. When a tool has been hands-on tested, its entry will say so, with the test date. Anything not labeled hands-on was researched from documentation only. We do not publish latency figures, deliverability rates, or benchmark scores we have not measured ourselves.

What we don't do

• We don't copy pricing from other comparison sites; if a number isn't on an official page, we mark it unverified or leave it out.
• We don't treat a vendor's marketing claim ("industry-leading", "fastest") as a finding.
• We don't invent a single numeric score. Rankings reflect fit for agent workloads — pricing transparency, free-tier usefulness, agent-specific features — and each entry explains its position in plain language.

How rankings change

A ranking changes when the underlying facts change: a vendor reprices, a free tier shrinks or grows, an agent-relevant feature ships or is deprecated, or a hands-on test contradicts the documentation. We re-check categories on a rolling basis and update the "checked" date on each page when we do. Vendors and readers can flag stale facts at listings@bestagentstack.com — include a link to the official page that proves the change and the correction is free and fast.

Sponsorship and editorial policy

BestAgentStack earns money from labeled sponsorships and featured placements. The rules are fixed and public:

• Sponsors never buy rank. Ordering within a comparison is editorial and is not for sale, at any price.
• Sponsored placements are visually separated and labeled "Sponsored" wherever they appear.
• A sponsor's verdict text is written the same way as every other verdict — including its limits.
• If a vendor pays us and also ranks first, it ranks first because of the checked facts, and the sponsorship label still shows.

This policy is the business model, not a disclaimer: a directory that sells rank stops being believed — by readers and by the AI systems that cite it — and belief is the asset.