Your code sends the page address, the policy, the purpose and the crawler name.
The two calls most integrations need
A check decides one page for one purpose. You use it when you want an answer for a particular address, or when you need the page's own tags and headers taken into account.
A site bundle gives one answer for a whole site, covering the purposes in your policy. You use it inside a crawl, where asking once per site is enough and speed matters.
How a check works
The API answers at once with a check number and tells you how many seconds to wait.
RightSignal reads the site politely. It fetches robots.txt first and keeps to the site's own limits.
Your code asks for the result and receives the decision with its receipt number.
A request looks like this.
POST /v1/gate/checks
Host: api.rightsignal.legal
Authorization: Bearer rs_test_your_key
Content-Type: application/json
{
"url": "https://www.example.in/news/markets-today",
"policy_id": "india-ai-training",
"purpose": "ai_training",
"crawler": "GPTBot"
}Click the scene for the next step
What a decision contains
- It contains the outcome and its group, which is collect, hold or stop. Your code can branch on the group and show the outcome to people.
- It contains the reason in plain words.
- It contains the quoted lines the outcome rests on, each with its source.
- It contains every condition and every evidence gap.
- It contains the profile, the purpose and the crawler the decision was made for.
- It contains the version of the RightSignal Algorithm, the time each source was read, the time of the decision and the time at which it stops being current.
- It says whether the AI reader and the policy-page assistant took part.
- It contains the seal and the receipt number.
No response carries a score, a probability or a percentage.
How the site bundle keeps a crawl fast
A crawler should not wait on a network call before every page. The site bundle is built so that it does not have to.
- Your crawler asks once for a site and receives the current decision for each purpose in your policy, with the crawler instructions sealed with it.
- The answer says how long it stays current. Your crawler caches it for that time and checks pages against it locally.
- When the answer expires, your crawler asks again. If nothing has changed, the API says so, and an unchanged bundle is not counted against your plan.
- If a site has no current decision, the API starts the checks and tells your crawler when to come back.
The first read of a site takes as long as a polite read takes, because RightSignal keeps to the site's own rate limits. Run your site list before the crawl, and the answers will be waiting when the crawl begins.
Test mode and live mode
Every account has both modes, and each key belongs to one of them.
- Test mode is for building. Decisions carry test seals that are clearly marked. The AI reader does not run, and each decision says so. Test usage is counted apart from live usage.
- Live mode is for evidence. Decisions are sealed with RightSignal's service key, and the AI reader and the policy-page assistant can take part.
The request and the response have the same shape in both modes, so code that works in test mode works in live mode.
Receipts in your pipeline
- Collect fetch
- Collect on conditions fetch, and keep the conditions
- Hold for review flag for a person
- Do not collect skip
Store the receipt number with every item you collect. It is the link between a row in your dataset and the evidence for it. Your code can fetch the sealed receipt by its number at any time. Anyone can verify a receipt with the standalone verifier, which runs offline and needs no connection to RightSignal. Export and keep your own copies of your receipts, so that your evidence does not depend on any one system.
Keys, limits and usage
- Keys are created in the app and shown once. Each key has a mode and a set of permissions, and you can rotate or revoke it.
- Each key has a rate limit. When you exceed it, the API tells you how long to wait.
- Each plan has monthly allowances. The usage and limits calls tell your code what remains.
- Every request is written to your organisation's audit log, which your owners can read in the app.
- Members of your organisation have roles, so the people who run checks are not all able to manage keys.
Building for a careful pipeline
RightSignal suggests three habits. Treat an unanswered check as a hold and do not fetch until it is answered. Act on the outcome group in code, and send holds to a person with the quoted lines. Keep the conditions with the data they apply to, so that a condition accepted at collection is still visible at training time.
What is not built yet
RightSignal has no software kit for a particular language, no plugin for a scraping framework and no tool for AI agents today. The API is the integration. Custom policies and registration of your own crawler's name are planned. The ready-made policies cover AI training under each profile and text and data mining under the EU profile. We will say so on this page when any of this changes.
Documentation
The full reference, a getting-started guide and a machine-readable description of the API will be published at /docs. Until then RightSignal sends them to every team that is given access. The API terms set out the rules for using the API.