Free Company Logo API: Any Domain to a 128 px Logo URL

Two different problems hide inside "find me this company's logo", and only one of them is hard. Fetching an image once you know the host is close to mechanical — every site maintained in the last decade publishes an icon, and there is a well-known convention for where it lives. Working out which host a brand name belongs to is the hard half, and it is the half that fails silently.

"Stripe" is a brand, not a host. So is "Delta" — which is an airline, a faucet manufacturer, and a dental insurer, depending on who is asking. So is "Apple", "Square", "Monzo" and every other name that reads as an English word. Resolving a name to the domain that company actually owns is where a logo list quietly goes wrong, and the failure is silent: nobody notices that row 14 of the customer wall is a different company with a similar name until that company's legal team does.

This tool runs both steps. A name is matched against a company index to find its domain; the domain is then converted into a logo endpoint. Enter a domain and the first step is skipped, which is why the two paths behave differently:

• Name path. Depends on the company being in the index. Returns a shortlist rather than one answer, on purpose — you are better placed to pick between "Delta Air Lines" and "Delta Faucet" than any matcher is.
• Domain path. Depends on nothing. The logo address is derived from the host you typed, so a domain registered last week returns an image just as reliably as a Fortune 500 one.

If you have the domain, use it. The name path is a convenience; the domain path is a guarantee.

The logo comes back at 128 pixels. That is not a limitation that will be lifted in a later version — it follows from where the image comes from.

An icon endpoint serves the mark a site publishes for browsers: the thing in the tab, the bookmark, the phone home screen. Sites publish it because browsers require it, which is exactly why this approach works for every domain instead of only for companies with a public brand kit. The trade is that you get the asset a site made for a 32-pixel tab, upscaled to a size that survives a list row.

That makes it right for:

• avatars and list rows
• comparison tables and pricing grids
• integration directories and customer walls
• favicon-scale UI elements

And wrong for:

• hero images and page banners
• slide decks projected at any real size
• anything printed
• anything you plan to recolor, crop tightly, or place on a contrasting background

There is no vector behind it and no larger size to request. If you need an SVG or a 2000-pixel PNG, the source is the company's own brand or press page, and you will be reading their usage terms anyway — see below.

The failure mode worth knowing

When a domain publishes no icon at all, the endpoint does not fail. It returns a generated placeholder — typically a letter tile — with a normal successful response.

This matters more in code than in the browser. A script that checks only whether the request succeeded will happily store a placeholder as a company's logo, and you will find out when the page ships. Parked domains, expired domains and domains behind a holding page all produce this. If you are pulling logos in a loop, look at the output before it goes live.

Each result gives you both: a URL you can reference directly, and a download link. They are not interchangeable, and the choice has consequences.

Hotlink when the logo lives in a running product. The URL resolves against the company's current icon at request time. Nothing is cached on our side, so a rebrand propagates to your page without anyone touching it. For a customer wall, an integrations page or an avatar fallback — surfaces that get stale precisely because nobody owns them — this is the whole argument. The counter-argument is real too: you have taken on a third-party request, which means it can be slow, blocked by a strict content-security policy, or unavailable while you are up.

Download when the logo is going into artwork. A slide deck, a mockup, a PDF, or anything a designer will edit needs a file that cannot change under you. The point of an asset in a pitch deck is that it looks the same the day you present it as it did the day you built it.

A reasonable middle path for production systems: fetch the URL once, store the file yourself, and keep the source URL alongside it so you can refresh on a schedule. You get stability and a way out of staleness, at the cost of writing the refresh job.

A logo on a page is an image, and images are among the most reliably broken things on the web. Across the 100,000+ sites our audit engine crawled between January and July 2026, images were a critical-failure category on 63.5% of them. The most common failure inside that category was missing alt text, on a median 19% of a site's images — it is the third most-failed individual rule we check.

Four things go wrong specifically with logos:

No alt text. A logo is one of the few images where the correct alt text is obvious — the company name, plus "logo" — and it still gets skipped, because logos are usually added in bulk by whoever built the grid rather than by whoever writes the copy. If a mark is purely decorative and the company name is already in adjacent text, an empty alt="" is the right answer; a missing attribute never is.

No dimensions. An image without width and height attributes gives the browser nothing to reserve space with, so the row jumps when the logo loads. A grid of twenty of them is a layout-shift problem that shows up in your Core Web Vitals rather than in your eyes. You know the size here — it is 128 by 128 — so there is no excuse for omitting it. Check the effect with the free website speed test.

Eager loading of below-the-fold marks. Customer walls sit near the bottom of a page. Twenty logos loading during the initial render compete with content the visitor can actually see. Lazy-load them.

Reaching for these as your own brand assets. Your own og:image, your Organization schema logo and your favicon should point at a real file you control, sized how you chose. Across the same audits, 19.4% of sites were missing an og:image entirely, and structured data was a failing category on 64.8% — most often a missing or invalid Organization block, which is where your own logo belongs. Fix the first with the meta tags checker and the second with the schema markup generator.

Finding a logo is a technical question. Using one is not, and the two get conflated constantly because the first is now trivial.

Company logos are trademarks. Public reachability is not permission — the mark being one HTTP request away says nothing about whether you may put it on a page. Most established companies publish brand guidelines, usually at /brand, /press or /legal/trademark, and those pages generally address exactly the uses this tool serves:

• Customer and client walls. Almost always requires permission, and often an approved asset rather than a scraped icon. Many companies are happy to grant it; the ask is the point.
• Integration and compatibility pages. Usually permitted under a nominative-use clause, frequently with conditions on sizing, clear space and not implying endorsement.
• Comparison tables. Generally the most defensible use — you are identifying a competitor to discuss them — but tone matters, and some guidelines restrict comparative contexts explicitly.
• Anything suggesting partnership or endorsement. Reliably the line that gets crossed.

Read the guidelines for any mark that will be seen by customers, and get written permission where it is asked for. This is a description of common practice, not legal advice; when there is money or risk attached, ask someone qualified.

Being clear about the boundary saves more time than any feature:

• You have a picture and need the company's name. That is image recognition, and it works in the opposite direction from this tool. A reverse image search is what you want.
• You need print or large-format artwork. Go to the company's brand or press page for a vector.
• You need your own logo somewhere. Use your own file. Never resolve your own domain here and ship the result.
• You have a list of 500 companies. This is a single-query lookup — one name or domain per search, with a download link on each row. There is no upload, no batch mode and no CSV export.
• You need a supported API with a contract. There is no public endpoint, key, quota or SLA. The stable thing you can build on is the returned URL, not a documented service.

If what you need is the mark for a company you can name, at interface scale, without an account — that is the case this was built for.