datareadyStart
Add dataready to your tool with one line, run in your app's folder (the lines for every tool are on the MCP page):
claude mcp add dataready -- npx -y dataready-mcp
It runs on your computer and keeps your app's data in .dataready/ in your project; from us it downloads only the personas and the know-how. Sign in with your email the first time. Then say map my app to get your app's profile. To test a public page (on the quick-check server, see the MCP page), say:
test my page https://your-site.example
That is it. Your coding agent gets a brief, opens the page, answers as the person, and gives you a report link plus a six-character access code. The same link and code arrive by email.
One person
https://mcp.dataready.si/mcp; Claude Code: claude mcp add --transport http dataready-check https://mcp.dataready.si/mcp (all tools on the MCP page).A single walk is the quickest check. Name the person, describe them, or let us choose:
test my pricing page as dr-0042 walk through my signup flow as a busy CFO on a phone does this landing page work for a first-time shopper?
Add a goal and steps when you care about a specific path:
test https://your-site.example as a skeptical founder, goal: start the free trial, steps: open pricing, compare plans, click Start
The brief tells your agent to land, read the first screen for five seconds, scroll, look for plan information, try the main button up to the point of real data or payment, and look for reasons to trust you. It records the steps it could not do instead of guessing.
Panels
A panel sends a varied group through the same page and combines what they said:
test my landing page with 20 skeptical EU founders run a panel of 12 on https://your-site.example for small shop owners
Panels take 3 to 50 people; 12 is the default. We pick a mix of types and markets so one kind of person does not dominate. If your tool can run several helpers at once, it walks them in parallel; otherwise one after another. Answers go back in batches, and the report is written when everyone is in, or when you say "finish" with at least three walks done.
Audiences
Describe your audience the way you would to a colleague. Words we understand include markets (EU, US, DACH, Nordics, Balkans…), roles (founder, CFO, marketer, student…), industries (SaaS, e-commerce, health…), devices (mobile, desktop) and temperament (skeptical, impatient, early adopter, bargain hunter). A few Slovenian, Croatian and German words work too.
Prefer precision? Ask for filters: market, industry, role, seniority, device, type, and 1-to-5 ranges for tech level, buying power, skepticism and patience. Or pick exact people from the directory and say "use dr-0042, dr-0107 and dr-1300".
Reading the report
Every number is out of 100 unless it says otherwise.
| Part | What it means |
|---|---|
| Overall and grade | A weighted mix: intent counts most, then copy, trust and whether it works. A is 90 and up, B 75, C 60, D 40, below that F. |
| Intent | How likely people are to sign up or buy today, lowered by blockers on the way. |
| Copy | Was it clear what this is and who it is for, was it convincing, did the headline land, how much jargon. |
| Trust | How much they trusted it, and which proof they found or missed. |
| It works | The share of checks and steps that actually worked. |
| Would act | Yes, maybe or no to your main button, as shares of the panel. |
| Segments | Intent by device, market, seniority, tech level and skepticism, weakest first. |
| Friction map | Where people got stuck, ordered by how bad and how common. |
| Objections and quotes | The reasons to say no, grouped, and the most and least convinced people in their own words. |
| Confidence | Low under 5 walks, medium under 20, high from 20. |
Answers that contradict themselves count half. If everyone says nearly the same thing, the report says so: it usually means the audience was too narrow.
Report pages are private. They open with the access code, stay up for seven days, and are not indexed by search engines.
Proposing an agent
Missing a kind of customer? Use the form on the agents page, or tell your coding agent "propose an agent" and it will ask you a few questions. A good proposal:
- describes a type of person, never a real individual or a brand;
- has concrete goals and objections ("will not enter a card before seeing the product");
- has no links, markup or instructions: it is a description, and we refuse text that tries to be anything else.
A proposal from the form first mails you a confirmation link: press it within 48 hours, or nothing happens and the proposal is deleted. A proposal through your coding agent skips that step, because you are already signed in. We review every confirmed proposal and email you the decision. Approved people get an id starting with dr-c and join the directory. You license the text under CC BY 4.0; your name is shown only if you ask for credit, your email never.
Limits
- 200 single briefs, 10 panels and 30 report pages an hour per account.
- Panels of 3 to 50 people; answers submitted in batches of up to 20.
- Text snapshots: 60 an hour, the same site once every 20 seconds, public pages only, up to 12,000 characters of text.
- Three proposals an hour from one network and five a day per email.
FAQ
My agent says it has no browser.
Say "use a snapshot" and the brief will carry the page's text. Interactive checks (does the button work?) are then marked as not tested.
Can it test pages behind a login?
Only with your agent's own browser, and only with test credentials you give it yourself. Never share real ones.
Where is the report?
Say "give me the report link". The link and code are also in your email.
They found problems. Now what?
Your agent can fix the page if you ask it to. If you want it made production-ready by people who do that for a living, see prodready.si.