Your website already says who buys it

Paste a URL. It reads the site, writes down your positioning, your buyers and the value props per persona — every line traceable to a sentence on a page — and hands you a scoring rubric as code you can run against your own list.

Read a websiteवापरण्यासाठी मोफत. काहीही इंस्टॉल करण्याची गरज नाही.

हे कसे काम करते

तीन पायऱ्या, आणि त्यातील एकही “कॉन्फिगर करा” नाही.

01
Give it a URL
One address, and anything you know that the site does not say — who has actually bought, which deal you lost and why. That goes in the brief and is treated as evidence, the same as a line on a page.
02
It reads the site, not the internet
Up to twelve pages on your own domain: pricing, product, customers, docs. Each one reduced to text and kept in the workspace, so next month's run can tell you what moved.
03
You get files, and one of them runs
Positioning, ICP, personas, value props, gaps — and score.py, which takes a CSV of companies and returns a score and a reason for each. Read it, argue with it, change the weights.

हे काय करू शकते

Every claim carries its source
Stated means the site says it, and the sentence is quoted with its URL. Derived means the agent concluded it, and it says what from. An ICP nobody can audit is indistinguishable from one that was invented, and this one is auditable line by line.
The scoring rubric is code
Not a paragraph describing how you might score leads — a Python file that reads a CSV and writes one back with a score and a reason per row. It runs in a Linux sandbox, it is yours, and it works on any list you can export.
It re-reads on a schedule
Positioning drifts and a document that was right in March is a liability in June. Ask it to check weekly and it compares against the pages it kept, rewriting only the sections that actually moved.

हे कसे सेट केले आहे

कार्यपद्धती, जेणेकरून साइन इन करण्यापूर्वी तुम्हाला काय मिळत आहे हे तुम्हाला माहीत असेल.

वातावरण: A Linux sandbox with a network, reaching your site through a logged proxy.
साधने: curl and a shell for the crawl, Python for the scorer, file tools for everything it writes.
मॉडेल: Claude Sonnet 5 by default, and you can change it mid-conversation.
सुरुवातीच्या फाइल्स: A brief to fill in. Drop a leads.csv beside it and the scorer will use it.
काय राहते: Every file stays in the workspace, including the pages it read, so a later run can diff them.

लोक ज्या गोष्टी विचारतात

“Read example.com and tell me who we are actually selling to”
“Which of the personas has the weakest evidence behind it?”
“Score leads.csv and show me the ten worst fits and why”
“Re-read the site every Monday and tell me if the pitch changed”

हे काय करणार नाही

यातील प्रत्येक मर्यादा आम्ही प्रत्यक्षात अनुभवली आहे, हा रोडमॅपचा भाग नाही.

It cannot find companies. There is no search tool and no company database behind it — it scores the list you bring, and the rubric it writes works in whatever tool holds your data.
It cannot read a site that renders entirely in the browser. A page with no server-rendered HTML arrives empty, and it will name the URL and say so rather than describe a page it never saw.
It will not write a number your site does not contain. No market sizes, no growth rates, no typical deal values. It has read one website and will not pretend otherwise.
It does not price your product and it does not write your outreach. A pricing page says what you charge, never what you should.

कोणते मॉडेल हे सर्वोत्तम करते

Measured over real runs of this agent, per model. A run counts as finished when it produced the answer on its own — nothing failed, nobody was asked to approve anything, and it did not run out of steps. Read all three columns together: a model that finishes fast by giving up scores badly on the first, and one that completes everything by grinding scores badly on the second.

Model	पूर्णतेचा दर	मध्यक खर्च	मध्यक वेळ	Runs
claude-sonnet-5anthropic	100%	$0.077	15s	16
claude-opus-5anthropic	83%	$0.541	2m 55s	6
kimi-k3moonshotai	100%	$0.463	6m 54s	5

पूर्णतेचा दर: तुम्हाला हस्तक्षेप न करता काम पूर्ण करणाऱ्या रन्स.
मध्यक खर्च: क्रेडिट्समध्ये, एका सामान्य रनचा खर्च किती येतो.
मध्यक वेळ: भिंतीवरील घड्याळ, पहिल्या संदेशापासून उत्तरापर्यंत.

Medians over the runs behind each row. A model appears once it has 3 runs on this agent, and the run count is shown so you can judge how much a figure rests on.