Getting tolled

This is the whole job: tell naulon who wrote each article and where their cut goes, then point the toll at your origin. Your pages don't change. Your readers don't notice. Agents that want to read your work start paying for it.

"Article" here means any page you gate: an essay, a whitepaper, a research report, a reference doc. If an agent can read it at a URL, the toll can price it.

What you provide

A payout wallet. One EVM address that receives USDC on your site's settlement network, which is set per site under Sites → Manage → Advanced → Settlement network and named there. This is the whole requirement. Set it and every page the toll covers earns into it, including pages with no author, which is most of the web: documentation, a changelog, a pricing page, a company blog that ships under one masthead. Nothing here needs a byline. With no wallet set anywhere, pages read free: there is nothing to pay into, so the toll lets the request through.
Your origin URL and what the toll covers. The toll needs to know where to proxy reads and which URLs are billable. Two scopes:
- Specific sections: list the URL prefixes that count as articles (e.g. /essays/, /posts/). Everything else passes through untouched.
- Entire site: every page tolls, including the root and single-segment pages like /about. Feeds, sitemaps, robots.txt, favicons and static assets (CSS, JS, images, fonts) always stay free, because agents need those to discover what's worth buying, and you can exclude sections you want open (say, /legal/).
With specific sections, tell us the shape of your permalinks too. One level means the article sits one step under the section, as in /blog/my-post. Nested means something sits in between, which is what dated permalinks look like: /blog/2026/09/my-post. Get this wrong on a dated site and every post published that month is identified as 2026, so one entry in your catalogue answers for all of them, or none of them matches and the whole month reads free.

Pick it before you build your catalogue if you can. The setting decides how an article is identified, and that identifier is what your catalogue, your free list and any licence a buyer is already holding are all written against. Changing it later re-identifies every article at once: entries that no longer match stop earning, pages you marked free start charging, and a buyer holding a licence for the old identifier pays again. Nothing is deleted and switching back restores the old identifiers, but the entries themselves are not rewritten, so if you do change it, go back through your catalogue afterwards. If you serve a credits endpoint, it will be called with the new path from that moment on.

One price for the whole site is the default, not the limit. An archive you wrote once and the investigation that took three months are not the same asset, so a site price can be overridden per section under Sites → Manage → What tolls → Pricing. A rule is a path pattern plus a price, a citation multiplier, or both; anything a rule does not cover keeps the site price.

The patterns are the ones robots.txt has used for thirty years: RFC 9309, the same dialect an RSL document already publishes and the same one a licence scope is written in. /papers/* covers everything under /papers/. Rules are allowed to overlap, because the ordinary thing to say is "everything under /papers/ costs more, except the preview", and that is two rules and one intent. The most specific match wins, exactly as it does in robots.txt, and the editor stores and shows them in that order so precedence is something you can read rather than something you have to work out. The same pattern written twice with two different prices is the one case refused, because that has no answer.

A rule cannot price a section at zero. Nothing below the settlement floor can move on-chain, so a free rule would save cleanly and then be unhonourable at payment time. To open a section, exclude it from what tolls, which is the setting that actually makes a path read free.
A credits endpoint, only if you want per-article control over who is paid. One read-only HTTP route on your domain that, given a slug, returns who gets paid and how the share splits. Serve it when some pages should read free, or when you pay named authors instead of the site. The toll calls it in place of your site wallet; skip it and the site wallet answers for everything.

It does not price anything; the response carries no amount. Prices come from the base toll and the rules above, and both apply the same whether or not you serve this endpoint. Full spec: the credits API.

Per-author wallets matter only here. Each author who wants their own share supplies one EVM address; a co-author with no wallet folds into the primary rather than leaving money unclaimed.

That's it. You do not run a wallet for the operator, you do not hold buyer funds, and you do not change how your site is built or hosted.

How a request flows

When something requests one of your articles, the toll decides in this order:

Is it a human? Browser-shaped requests pass straight through, free. The classifier biases toward "human": if it's unsure, the read is free. You never risk charging a person.
Is it an agent, and is this article gated? The toll asks your credits API about the slug.
- If your API returns 404, the article is free. 404 is the deliberate "don't toll this one" signal. Use it for anything you want open.
- If your API returns credits, the toll quotes a price and answers the agent with 402 Payment Required.
The agent pays. It signs a payment for the quoted amount. The fee settles on-chain, buyer → author, against the wallets your credits API named. The toll then serves the article and hands back a short-lived license so the agent can re-read the same piece without paying again.

What a fee is

A price per read, set per article, plus a higher multiplier when the read is a citation: an agent pulling your text to ground an answer it will repeat elsewhere. Citations are worth more than a glance, so they cost more. All amounts are whole units of micro-USDC; there is no floating-point money anywhere in the math, so splits between co-authors are exact.

Co-authors and splits

An article can credit more than one author, each with a relative weight. The fee divides by weight, and each author's share settles to their own wallet in the same transaction, not to a pool that pays out later. If a co-author has no wallet, their weight folds into the primary author rather than leaving money unclaimed. You decide the weights; the toll does the arithmetic.

Authors the toll discovered for you

If you let naulon crawl your site instead of serving a credits endpoint, the bylines it finds become a roster of their own under Catalog → Authors, one per site. A crawled byline is just a string on a page, so the roster is how it becomes a person who can be paid.

Each row sits in one of three states, and moves in one direction. Crawled is what the crawl found. Invited is after you send that byline an invite: the crawl can read a name off a page but not an email address, so you supply it. Claimed is once they have accepted and set their own wallet, from which point their share settles to them directly, exactly as if you had named them in a credits endpoint.

You can also set a fallback payout wallet against a byline that has not claimed yet, so its share has somewhere to go in the meantime rather than folding into the primary author. Only sites naulon crawls have a roster; a site with its own credits endpoint names its authors there instead, and that is the page for it.

What the person on the other end sees is For authors; send them that.

Going live

Stand up your credits endpoint and confirm it returns the right shape for a gated slug and a clean 404 for a free one. See the credits API.
Add wallet addresses for the authors who want to earn.
Give us your origin URL and article path prefixes; we route your Host to your configuration.
Watch a real agent read a gated article, get a 402, pay, and read. The author's wallet receives the fee directly. See settlement for what lands and where.