FAQ

Do I need to know how to code?

No. Describe the data you want, correct what Ray9 proposes, and schedule it. Selectors, proxies, headers, and execution routes are there when you want to tune them, and hidden when you do not.

If you would rather work in code, the same Flows run from the API, the CLI, and MCP — against the same model, with the same rules.

What is a credit?

Ray9's metered unit of work. Cheap routes cost fewer credits than expensive ones — a direct fetch costs less than a stealth browser on a residential exit, because it costs less to perform.

Ray9 shows an estimated range before a run and the actual usage afterwards, reserves a maximum before dispatching anything, and stops cleanly at your budget rather than running up a bill. See Credits.

What happens when a target blocks the run?

You get a named failure — target_blocked, auth_required, rate_limited — with the routes that are still eligible and the recommended next action.

Ray9 will not silently relax a requirement you set, such as a geography or a logged-in session, to make a run succeed on a cheaper route. A run that quietly returned the wrong country's prices would be worse than a failed one, because you would believe it. See Execution routes.

What is Ray9 allowed to scrape?

Defaults favour public, logged-out, non-sensitive data with conservative rate limits, and target policy is evaluated before any credits are spent.

Ray9 does not build features whose purpose is bypassing logins, paywalls, or access controls. Authenticated jobs require your explicit action and come with isolated credential scopes, encryption, short retention, and an audit trail.

You remain responsible for the data you collect and how you use it — and none of this is legal advice.

How is this different from a scraping API?

A scraping API hands back a page. Ray9 owns the lifecycle around it: a versioned definition, a schema it validated, a dataset with observation history, a schedule, change monitoring, signed delivery, and a credit ledger — reachable from a dashboard, an API, a CLI, an MCP client, or an agent, all against the same model.

The difference shows up in month three, when the page changes and you need to know which of the last ninety runs the bad values came from.

Can I run a Flow in my own browser?

Not yet. A local browser agent — for jobs that must run on your device with your session — is on the roadmap, and the product is designed around an explicit trust boundary for it. Everything shipping today runs in the cloud.

How do I get the data out?

Filter and export from the dashboard as JSON or CSV, page through it on the public API, receive it as signed webhooks with replay protection and bounded retries, or pull it from an MCP client or agent.

Every record carries its source URL, Flow version, run, and observation time. See Datasets.

How do I know the extracted data is actually right?

You look at it before you publish. The builder shows sample records beside the captured page, with every field highlighted against the region it came from, and it flags the fields it is unsure about rather than quietly guessing.

After that, validation is structural: types, required fields, ranges, expected record counts, duplicate rules. HTTP 200 is never treated as success — a challenge page, a consent wall, and the wrong locale all return 200, and Ray9 classifies each of them as a failure. See Monitoring for how the same idea applies to changes.

What happens if a run only half works?

It is reported as partial, and the valid records are kept.

A partial result is never promoted to complete, and it is never thrown away. You see the valid, invalid, and duplicate counts separately, along with the evidence, and the run tells you what stopped it.

Can I stop it from spending money?

Yes, at several levels. Estimates are shown before a run. A maximum is reserved before dispatch, so a run cannot exceed it. Organizations can set per-run, daily, and monthly budgets. Schedules can carry their own budget and per-run caps.

You can also opt out of fallback routes that increase cost — if you would rather a run failed than escalated to an expensive route, Ray9 will fail it.

What happens when the page changes and my Flow breaks?

You get a classified failure — usually extraction_failed — rather than a run that silently returns nothing and a dataset that quietly stops growing.

Because the capture is retained, you can usually repair the field mapping and re-test against the stored evidence without fetching the page again. Fix the draft, publish a new version, and decide whether to promote your schedules to it. Past runs keep pointing at the version that actually produced them.

Can my team share an organization?

Yes. An organization has roles — owner, admin, builder, operator, viewer — that separate who can manage billing and members, who can build and publish Flows, who can run and schedule them, and who can only look.

Does Ray9 keep the raw pages?

It keeps evidence: the captured HTML, screenshots, and traces behind each attempt, with configurable retention. That is what makes a value in your dataset traceable months later, and what lets you repair an extraction without refetching.

Evidence retention is separate from record retention, because artifacts are much larger and usually needed for a shorter time. Both are subject to your organization's retention and deletion rules.

Is the pricing final?

No. Pricing is provisional while Ray9 is in early access. See the pricing section for current tiers, and talk to us before you build a budget around them.