Frequently asked questions

JupiterFetch is a brand API. Send it a company's domain and it returns that company's brand as one JSON document: logos, colors, fonts, images, social links and page metadata, read from the company's live website. Every image comes back as a link on our CDN, ready to use.

One JSON document per domain. It has the brand's name and description, its logos (the primary mark first, then alternates tagged by kind and theme), colors with roles such as primary and background plus a ranked palette, the fonts its text is set in, its share image and X card image, its website and social profile links, page metadata, and a quality score saying how much of that we found. The API reference lists every field.

Yes. The box on the home page runs a real lookup on any domain you type and shows the profile it returns, with no account. Each visitor gets 5 tries an hour.

Sign up and verify your email, and your account gets 100 free credits. Create a key on the API & MCP page, then follow the quick start: one GET request with your key in the Authorization header.

No. The 100 free credits arrive when you verify your email, and no card is asked for. You add a card only when you buy credits.

No. Every key is live, and every call that returns a profile uses a credit. The 100 free credits are there so you can try it on real domains before you pay.

Yes. Send example.com or a full URL such as https://example.com/pricing. We always read the site's home page, so the path makes no difference. www.example.com and example.com are separate lookups, and so is a subdomain such as shop.example.com, because each can carry a different brand. If the form you send does not respond, we try the other one.

No. The API is one GET request with a bearer key, so any HTTP client works, and the response is plain JSON described in the API reference. Once you are signed in, the API & MCP page gives you a prompt to paste into a coding agent so it can write the integration for you.

Yes. Add https://mcp.jupiterfetch.com as a connector in Claude, ChatGPT or any other MCP client, sign in with your JupiterFetch account, and ask for a company's brand. JupiterFetch is listed in Claude's connector directory. Each successful lookup spends one credit, the same as the API. The set-up steps cover each app.

$0.01 per successful call, paid for with prepaid credits: one credit is one call. A new account gets 100 free credits once its email is verified. After that you buy credits when you need them, from $10.00 for 1,000 up to $2,000.00 at a time. There is no subscription. Pricing has the details.

Every call that returns a brand profile, a 200, uses one credit. That includes a profile served from our cache, a thin profile with a low quality score, and an older copy we return marked stale when a fresh read fails. A call that ends in an error uses none.

No. Only a 200 with a brand profile spends a credit. A 504 timeout, a domain that does not resolve, or a 424 when a site blocks us costs nothing, even when the 424 carries a partial profile. A 200 always costs one credit, including a thin profile and an older copy marked stale.

Until you spend them. Credits never expire, so they stay on your account however long you leave them.

No. You buy credits when you want them, and they stay on your account until you spend them. Nothing is charged unless you buy credits or switch on auto top-up, which is off until you turn it on.

The API answers 402 quota_exceeded, and the X-Credits-Remaining header reads 0. Nothing is charged automatically unless you have switched on auto top-up. Buy credits on the Billing page and calls work again straight away.

It is off until you switch it on in Billing, and it needs the card from an earlier purchase. When it is on and your balance drops below 500 credits, we charge that card a set amount, $10.00 unless you change it. A monthly cap, $100.00 unless you change it, limits what it can charge in a calendar month (UTC). When the next top-up would pass the cap, auto top-up pauses until the month ends. If your card is declined, it pauses and we email you; update the card, then press Resume on the Billing page.

By card, through Stripe's checkout, in US dollars. On the Billing page, buy any whole-dollar amount from $10.00 to $2,000.00. Stripe emails you a receipt for each payment, and the Billing page lists them. To buy more than $2,000.00 at once, message us.

No. Credits are non-refundable, as the terms say. They also never expire, so anything you do not spend now stays on your account.

We read what the site itself shows and declares, so a profile is only as complete as the site. The usual causes are a site that blocks automated requests, or one that declares little about itself: no clear logo, few colors in its stylesheets. Every profile carries a quality score that says how much we found, so your code can decide what to do with a thin one. If a profile looks wrong for a site you care about, message us with the domain.

quality.score runs from 0 to 100 and measures how much of the profile we found, not how good the brand is. A score near 100 means we found a real logo, several colors with a clear primary, the site's fonts, its name and its links. A low score means some of those are missing, and it is a normal result for a sparse site, not an error. The API reference lists the coverage flags behind it.

Usually no more than seven days old. We keep each profile for seven days, and the first request after that reads the site again. If that read fails or runs out of time, you get the previous copy instead of an error, marked stale in the Cache-Status header, and it can be up to 14 days old. If the site blocked us, you get a 424 instead. You cannot force an earlier re-read. If a company has just changed its brand and you need the new one now, message us.

Most profiles come back in a few seconds, and one already in our cache comes back at once. A site we have to load in a full browser takes longer. Each call waits up to 20 seconds by default; set wait to as much as 60000 (milliseconds) if you would rather wait longer than get a timeout.

You get a 504 fetch_timeout, which costs nothing. The fetch keeps running after we answer and its result goes into our cache, so asking for the same domain a little later usually returns the profile at once. If we still hold an older copy, we return that instead of the 504, marked stale.

You get a 424 site_blocked problem document, which costs nothing, rather than a confident-looking empty profile. If the page partly loaded, the 424 carries what it did yield. Some sites refuse every automated request, and for those we cannot return a full profile.

Yes. Logos, colors, fonts and links do not depend on language, and the name and description come back in whatever language the site uses. meta.language reports the language the page declares.

Yes. Every image link is a plain HTTPS URL on our CDN, so it works as it is in an <img> tag or a CSS background. Each link works for at least 30 days after we return it. After that it may stop, so if you need an image for longer, download it within that window and serve your own copy.

Each logo and image is a PNG, JPEG, WebP or GIF on our CDN. When a site's logo is an SVG, we also return a cleaned copy of the SVG at vectorUrl, with scripts and other unsafe parts removed, and a PNG of it, 2,048 pixels on its longest side (up to 4,096 for a very wide or tall mark).

You can download and keep a copy of anything we return, and we add no attribution requirement. The logos themselves stay the property of their owners, and how you use them is your responsibility.

10 requests a minute per account until your first purchase, then 120 a minute. The limit belongs to the account: every key on it and every connected AI app share it, so a second key does not raise it. Past the limit you get a 429 rate_limited with a Retry-After header saying how many seconds to wait, and it costs nothing. A 429 can also mean we are briefly at capacity; the same header tells you when to retry. In an AI app, the tool's reply says how long to wait instead.

No. Each call looks up one domain. To look up a list, send one call per domain and stay within your rate limit; a 429 tells you when to slow down.

No. Every call carries your secret key in its Authorization header, and a key in browser code can be read by anyone who opens the page. Call the API from your server and pass the result to your front end. The images are the exception: their CDN links are public and safe to use in a browser.

There are no team accounts. One account can hold up to 10 API keys, so you can give each service or teammate its own key and revoke one without touching the others. Every key spends from the same credit balance and shares the account's rate limit.

Use Disconnect next to the app on the API & MCP page. That ends its access at once and does not touch your API keys. Removing the connector inside the app is not the same: the access it was given stays valid until it lapses, which can take up to 30 days.

Message us and ask for it to be deleted. Deleting it removes your account data and your usage history; the privacy policy says what is kept. Unused credits are not refunded.

No. We do not offer an uptime guarantee or a public status page, and the terms say we do not promise uninterrupted service. A call that fails on our side costs nothing.

For each call we record the domain you asked about, the result, whether it used a credit, how long it took and which key or app made it. Your usage page shows your recent requests, and all of it is deleted with your account. Brand profiles are about companies, not about you: one cached profile answers every customer who asks for that domain. The privacy policy has the full list.

One page load of the domain a customer asked about, plus the files that page references: stylesheets, fonts, logo and share images, the favicon and the web manifest. It is not a crawl: we do not discover or queue sites on our own. We keep each profile for seven days and answer repeat requests from that copy.