How we review

Who reviews

Pages are written and checked by the Evals Review editorial team, which takes editorial responsibility for everything published on evalsreview.com. Corrections go to the same team through the contact page or the submit form.

What we check, and how

• Tools (12 LLM evaluation platforms, 26 AI agent tools): free tier, entry price, pricing meter, licence, self-hosting, API, MCP server, compliance claims and data retention, each read on the vendor's own pricing page, docs or official repository. Every field links to its source.
• Evals (79 benchmarks, datasets and rubrics): who made it, what it tests, licence and last update, taken from the paper, repository or official page.
• “not verified”: if we can't confirm something on an official page, the page says “not verified”. That isn't the same as “no”.
• What we don't do: we don't run hands-on benchmarks of every tool and we don't claim to have tested a product unless the page describes how the test was run. We don't copy vendor marketing; summaries are our own words.

How lists and comparisons are ordered

• Tool lists and “best X for Y” pages are alphabetical and every tool gets the same fields. They aren't ranked and carry no star ratings or scores.
• Head-to-head pages don't name a winner. The verdict says when each option fits, based on the sourced facts in the table.
• Short answers at the top of a page summarise those same sourced facts, for example which tools publish self-hostable code or the lowest published entry price.
• Commission and sponsorship don't affect order, inclusion, verdicts or facts.

User opinions and reviews

• We show no user ratings. “What real users say” sections quote public posts (for example Hacker News and GitHub issues) word for word, one sentence or less, with a link and date. We check each quote against its source and include critical posts as well as positive ones. We can't confirm that each poster used the product, and we never use paid or incentivized reviews. Last checked 2026-10-18.
• Eval pages can carry reviews from people who sign in with GitHub, one per account per eval. We can't confirm each reviewer used the eval; we remove reviews that break our rules, never just for being negative, and we never write, buy or edit reviews.

How we earn money

• Affiliate links: some tool pages have an affiliate “Visit” button. If you buy through it, we earn a commission at no extra cost to you. Those pages show an “Ad / affiliate disclosure” at the top, the button is marked “Ad: affiliate link”, and the link uses rel="sponsored nofollow". Pages with affiliate links now: Firecrawl. Details: affiliate disclosure.
• Sponsorship: vendors can buy labelled placements (see Sponsor). Sponsored items are marked “Sponsored” and never change order, verdicts or facts.
• Source and citation links always go to the vendor's or creator's own pages, never through an affiliate link. Our CSV and JSON data contain no affiliate links.

Update policy

• Every price shows the date it was checked. Tool prices were last checked 2026-10-17 (LLM evaluation tools) and 2026-10-17 (AI agent tools), so they are “as of October 2026”. Always check the vendor site before you buy.
• We re-check prices at least every 90 days. If we can't verify a price again, we remove the number rather than leave it stale.
• We fix factual errors when we're told about them with a source link, whether or not the vendor is a sponsor or affiliate.

Use of AI

We use AI assistance for research, drafting and data checks. Every fact on a page comes from the linked source, and pages are checked against those sources before they're published. The Evals Review editorial team is responsible for what we publish. We don't use AI-generated reviews, testimonials or people, and AI-narrated videos say so where required.