agentic

Every listing carries a 0-36 score built from nine dimensions, each rated 0-4. Action capability checks whether a tool touches APIs, files, or other systems; autonomy checks whether it picks its own next step; planning, adaptation, state, reliability, interoperability, safety, and operator sovereignty make up the rest. The site states plainly that a higher number means more independence, not higher quality, so a low score is not a knock. That framing keeps the directory useful when you want a simple assist instead of a fully autonomous operator.

A raw number is hard to shop with, so the directory sorts every score into one of six levels. A Reactive Tool at 0-5 answers prompts but never acts, a Guided Assistant at 6-12 runs steps you hand it, and it climbs through Adaptive Collaborator, Domain Specialist, and Autonomous Operator up to a Strategic Agent at 32-36. Vendors cannot buy their way up this scale. Reading the level first, then the score, tells you fast whether a tool matches how much control you want to hand over.

The home page asks one question: what are you trying to do? Type a sentence and you get a single honest pick, with a note that the site will say when the answer isn't its own. That runs alongside category browsing and an AI search for people who already know the shape of what they need. It is aimed at cutting the list-down time, since a one-line answer fits someone who wants a decision, not another dashboard.

A tool page splits the total into the nine dimensions and shows where the evidence was strong and where it was thin. It also labels confidence, so a tool with limited published detail is marked as scored conservatively. This is the part that separates a review from a rating: you can see that a tool plans well but has no audit trail, or handles errors but locks you into one model. Each listing shows when it was last evaluated, so an old score is visible as an old score.

Beyond the raw listings, the directory publishes ranked guides such as the best AI coding agents in 2026, the best free ones, and the best for research and analysis. Each guide says how many tools it covers and that placement is not for sale, and rankings come from the rubric score alone. The guides are where the directory does its own arguing, comparing tools head-to-head rather than leaving a flat list for you to sort.

Tool makers can claim a listing, add screenshots, video, and pricing, and answer user questions. Submitting a tool is free, and a paid tier adds screenshots, longer analytics, and priority placement. Sponsor slots sit at the top of a category for $29 to $129 a month, clearly labelled and kept outside the ranked grid. The directory says the score itself cannot be bought and re-evaluates every listing regularly, which is the claim to check when a sponsored card sits above a lower-scored tool.