The operator never pays. The rails and the carriers do.
We rate every operator in the agent economy from the outside, unasked, the way a credit bureau rates a borrower. That rating is free to the operator permanently — not as a trial, not as a teaser. It is the distribution instrument, and charging for it would buy us nothing we measured anyone willing to pay for. The money is priced where a budget provably already exists: the rails fighting over these operators, and the carriers trying to underwrite them.
6,753 operators rated · 15 of them with a dated critical finding · facts as of 2026-11-16
Pricing doctrine. Three different buyers, three different willingness to pay, one shared dated series underneath. The operator-facing product is FREE on purpose: our own demand census measured that price explains ~0% of demand on this box and that 0 of 155 checkout sessions were ever a real purchase attempt — the constraint is traffic, so charging the operator buys nothing and costs us the distribution instrument. The money is priced where a budget provably already exists: rail/platform BD and insurance underwriting.
Operator dossier
Free — permanently
Buyer: The rated operator
Every one of the rated operators has a page with their name on it: grade, rank, the five dimensions against the cohort, rail position, surface stability, and any dated finding we hold.
Grade + percentile rank against the whole rated population
Each dimension vs the cohort median and top decile
Rail position: x402, A2A, or neither — probed, not assumed
Any dated surface-change finding on that host
Dispute route: recompute the source yourself
Why this number: Deliberately $0, not unpriced. Measured on this box: price explains ~0% of demand, and 0 of 155 checkout sessions were a real attempt. A paid operator report converts nobody and kills the one instrument that generates named counterparties. The dossier IS the distribution.
Buyer: A payment rail, gateway, or registry with a BD team
Which operators arrived on your rail, which defected, which high-value nodes neither rail has captured — dated, re-derivable, and impossible to backfill.
Dated operator flow: arrived / left / retained, by name
The unclaimed high-value nodes ranked by node score
Your base's concentration and defection fragility (HHI, top-1/top-3 share)
Rail-vs-rail overlap: who bridges both, who is contested
Honest scope: only x402 and A2A are observable at operator level; closed card rails are reported UNOBSERVABLE, never zero
Why this number: Anchored on StillOS's own existing Marketplace Trust Layer price (integration from $2,500, platform plans from $1,000/mo) — internal consistency, not a new invented number. Justified by exclusivity, not by scale: the dated defection series cannot be bought from anyone, including us, retroactively. A rail's BD team cannot currently answer 'who left us last month' at all.
Buyer: A cyber carrier or MGA writing agentic/supply-chain risk
The accumulation exposure across the agent economy: which single dependencies sit under thousands of operators, how concentrated the hosting is, and how that moved quarter over quarter.
Shared-chokepoint accumulation: one dependency, N operators downstream
Interval movement from a dated series that began 2026-09-25 (the stable watch-set regime). The first full quarter-over-quarter delta lands 2026-12-25; until then you receive the dated baseline plus every interval delta since. Nobody can backfill this series, including us — which is the reason the start date matters.
Every figure sealed to the notary chain and re-derivable by your own analysts
Why this number: Anchored on StillOS's own existing enterprise floor (from $2,500/mo after qualification = $30k/yr); set at $24k/yr as four quarterly deliveries. Justified by the missing input, not by page count: a carrier cannot price agentic risk without an accumulation figure, and accumulation requires a dated before-picture of shared dependencies — which is the one thing a point-in-time scanner structurally cannot produce.
Bulk/dataset pulls (coverage, export, feeds) are paid per pull over x402 on Base — the agent pays and reads, no account. Per-entity lookups and the index are FREE PERMANENTLY; deep per-entity pages need a free key (identity, not payment) because a 23,639-IP clone defeats any per-IP counter.
HTTP 402 challenge on bulk paths, settled on-chain before the handler runs
Free key for deep per-entity pages: POST /ratings/key with {email} — no payment, ever
A deterministic 10% sample of per-entity pages stays open with no key at all
Index, methodology, sitemaps, /.well-known/* and the agent card: free forever, never gated
Why this number: CORRECTED 2026-11-01 (Marcus: "is anything still being advertised for free and not complete loops that require payment from external actors"). This tier advertised "$0.05 / call" and "See the live 402" in both the HTML board and the machine-readable offer.json, and NO /ratings route has ever charged $0.05 — /ratings/node, /ratings/offer.json and /ratings/floor all answer 200 free, and the only 402 on this host is the corpus gate's BULK tier at $0.25. The $0.05 was the NOTARY's measured settlement price on different endpoints, borrowed onto this board, so the "EVIDENCE-BACKED / observed" claim was about another product. Now read from what the gate actually enforces: core/corpus_access_gate.cjs returns PAYMENT_REQUIRED price_usd 0.25 on a bulk path and KEY_REQUIRED (no price) on a deep path. A false advertised price is the exact divergence we grade other vendors on.
One email, nothing else. No card, no account, no automated follow-up — this writes a line to a lead file a human reads.
Every number on this page recomputes from the sealed node board · dispute route: /notary/verify · an unobserved node is NOT attested safe, only unobserved · we rate operators who never asked to be rated, and we publish the method, not just the grade.