How to read scored.tools, how a score is made, and how to get the data out. If something here does not match what you see on the site, write tohello@scored.tools.
- Reading reviews
- How a score works
- Finding and comparing tools
- Feeds and the newsletter
- The public data API
- If you make one of these tools
Reading reviews
Every review, score, comparison and blog post is free to read. There is no account, no sign-up and no paywall, and no limit on how much you read. What readers pay is set out on the pricing page, and it is nothing.
Each tool has one review page at /tools/<name>/. It shows the overall score, the five criterion scores with the evidence and source link behind each one, the tool's published plans and prices, pros and cons, and a verdict. The page is dated, and a correction or an update says what changed and when.
Some links to tools are affiliate links, which can earn us a commission at no cost to you. Paid placements are labelled Sponsored where you can see them. Neither changes a score. The rules are in the editorial policy, and every payment is listed in the public ledger.
How a score works
Every tool is scored out of 10 on the same five criteria:
- Quality of output. How good the thing the tool produces is, judged on evidence of results.
- Ease of use. How quickly a new user gets to a first useful result.
- Pricing value. What you pay for the result, including any paid service the tool needs, set against comparable tools and the free path.
- API and integration quality. A documented API or SDK, integrations, import and export.
- Problem fit. Whether the tool does what its own homepage says it does.
The rules that apply to every tool:
- Each criterion is a whole number from 0 to 10, scored against fixed anchors, and carries a one-line evidence note and the URL of its main source.
- The evidence is the tool's homepage, its pricing page and its documentation page, its GitHub repository where it has one, and independent sources such as published benchmarks, reviews and reported problems. Marketing adjectives are not evidence.
- Without an independent source, quality of output and problem fit cannot score above 6.
- If a tool does not publish its prices, pricing value cannot score above 4.
- A criterion that does not apply to the kind of product, such as an API for a book or a course, or one with no evidence at all, is shown as n/a with the reason and left out of the average. A software product with no API is not n/a. It scores low.
- A tool needs at least three scored criteria to get a rating. The rating is the mean of the scored criteria to one decimal, rounded half up, and it is computed by code.
- Whether a vendor paid us is not part of the evidence. A paid review goes through the same procedure and is published whatever the score is.
We do not test every tool hands-on. A review labelled Tested is the exception, and its page carries a dated test log. The FAQ has more on scores and methodology.
Finding and comparing tools
- Search looks through the tool reviews and blog posts. Press Ctrl K on any page to open it.
- The tools directory lists all 571 scored tools. Each of the 13 category pages lists every rated tool in that category in score order, and the order is the score and nothing else.
- Compare puts two tools side by side on features, pricing and scores.
- The blog has reviews, comparisons and roundups.
Feeds and the newsletter
/rss.xml: the blog, newest posts first./sponsored-rss.xml: paid, labelled sponsored pages only, kept out of the main feed./llms.txt: a plain-text map of the editorial pages for AI assistants.- The newsletter sign-up is at the bottom of every page.
The public data API
Every score on the site is also published as JSON. It is read-only, free, and needs no key or account. The files are rebuilt each time the site is published, so they always match the review pages.
| Endpoint | What it returns |
|---|---|
GET /api/tools.json | Every scored tool (571 today), one row each, highest rating first. |
GET /api/tools/<slug>.json | One tool's full public record: the five criteria with score, evidence and source, pricing plans, features, pros, cons, verdict, competitors, dates and any published correction. |
Each row of tools.json has these fields:
| Field | Meaning |
|---|---|
slug | The tool's id on this site. Its record is at /api/tools/<slug>.json. |
name | The tool's name. |
url | The tool's own website. Never an affiliate link. |
category | Its main category, such as coding or writing. |
rating | The overall score out of 10, the mean of the scored criteria. |
scored_on | The date it was last scored (YYYY-MM-DD). |
rubric_version | The version of the rubric that scored it. |
page | The review page on scored.tools. |
api | The URL of the tool's full record. |
In a tool's record, criteria has five keys: output_quality, ease_of_use, pricing_value, api_integration and problem_fit. Each has a score (0 to 10), an evidence note and a source URL. An n/a criterion has score: null, na: true and the reason in evidence.
curl
# Every tool, one row each
curl -s https://scored.tools/api/tools.json
# One tool's full record
curl -s https://scored.tools/api/tools/hugging-face-transformers.jsonJavaScript
const res = await fetch("https://scored.tools/api/tools.json");
const { tools } = await res.json();
// The ten highest-rated coding tools
const coding = tools.filter((t) => t.category === "coding").slice(0, 10);
for (const t of coding) console.log(t.rating, t.name, t.page);
// The evidence behind one tool's score
const tool = await (await fetch("https://scored.tools/api/tools/hugging-face-transformers.json")).json();
console.log(tool.criteria.pricing_value);Python
import json
from urllib.request import urlopen
with urlopen("https://scored.tools/api/tools.json") as r:
index = json.load(r)
print(index["count"], "tools")
with urlopen("https://scored.tools/api/tools/hugging-face-transformers.json") as r:
tool = json.load(r)
for name, c in tool["criteria"].items():
print(name, "n/a" if c["na"] else c["score"], "-", c["evidence"])The files are sent with Access-Control-Allow-Origin: *, so a page on another site can read them with fetch.
Using the data. The text and scores belong to scored.tools, under section 6 of our terms: you may quote short extracts, such as a tool's score, with a link back to its review page (the page field). For anything larger, write to hello@scored.tools.
If you make one of these tools
These are free for every vendor, and none of them changes a score:
- Claim your page. Prove you run the tool and confirm the facts we list.
- Flag a fact. If an archived copy shows the fact was wrong on the date the page was published, we fix it free. A fact that changed later is updated free on our scheduled re-check cycle, with no guaranteed date.
- Reply to your review with a labelled vendor response below the verdict, and show your score badge on your own site. Both are explained under free for every vendor.
We don't take review requests, and our editors choose which tools to review. Paid options such as a Priority Review buy speed or a labelled placement, never a score. Every price is on the pricing page and the advertise page. Agencies and consultants who refer vendors can read about the invite-only partner program.