Verdict Row — fleet status

Generated 2026-10-26, live checks at 19:37–19:38 UTC; updated 2026-10-26 ~21:05 UTC for the Fleet Pass / Webhook Watch launch (StallRow paid routes 13 → 14, health re-checked 200 for all five services in the client-#0 demo pass).

Generated from the fleet's own watch state and live /health + /.well-known/x402 reads; if a number here disagrees with the chain or the catalogues, they win.

Services

23 paid routes live across 5 services (the 21st, StallRow /v1/fleet-pass, launched 2026-10-26 evening; its catalogue seeds are pending a wallet top-up — the route itself is live and its 402 is verified. Routes 22–23 — Verdict Index and Orchestrate — shipped later the same night; their seeds are on the same top-up list). Hosting: Railway Hobby plan. Payments: x402 v2 exact, USDC on Base, payTo 0x097f9685D1e0752bF8B02d61E5Ce9be7eA0FBA7F (public in every 402 challenge). Catalogue record at the 2026-10-26 integration watch: CDP Bazaar merchant total 19 and PayAI listing for the fleet's first 19 routes; the 20th route (LockSmith /v1/audit) has settled CDP + PayAI seeds and x402scan-verified ownership, with the CDP merchant count moving 19 → 20 once indexing catches up. All five services carry x402scan-verified ownership.

Proof

• Settled scout record (2026-10-26): 35 purchases / 35 delivered / 19 endpoints, all stamped as_advertised after the $2 whole-fleet re-test plus paid single-endpoint re-tests.
• Security audit 2026-10-25: no critical findings; one HIGH (SSRF in RouteVerdict / ListSmith URL fetching) fixed and deployed the same evening; 411 tests across the four audited services at audit time.
• Failure honesty rule: an upstream failure returns 502 and the payment does not settle — nobody is charged for undelivered data.
• Recheck store is persistent across redeploys (StallRow /health reports recheck_store: persistent): packs and watches survive deploys.

Accuracy Ledger — Token Verdict's track record, scored in public

Every Token Verdict verdict is now recorded and scored against what the token actually did afterward, under rules fixed in writing before the first scoring run (accuracy-ledger/METHODOLOGY.md in the loop-kits repo: mature at +7d and +30d; collapsed = pair gone or liquidity under $1,000; SOLID correct iff not collapsed and the move is better than −30%; CAUTION never scored for or against).

• Status: tracking started 2026-10-26. Records tracked: 6 — all SOLID, two tokens (WETH, AERO on Base), 2026-10-25 → 2026-10-26, all sourced from verdict-loop's public live log (labelled self-funded demo runs, because they are).
• Records mature for scoring: 0. The first +7d evaluations land 2026-11-01. No hit rate is published — a percentage over zero mature records would be a fabricated number.
• Verdicts served before the recorder shipped were never logged by the service and are unrecoverable; the ledger starts from what exists and says so, rather than reconstructing a prettier history. From 2026-10-26 the service appends every served verdict to its own ledger, so coverage grows with real traffic.

Verdict Row Verified — badge

Our own stalls were judged first, in public, under the criteria in badge/CRITERIA.md: Token Verdict, RouteVerdict, ListSmith and LockSmith are badged (battery 2026-10-26: every route's live 402 exact-matches its manifest, 22/22). StallRow is not — its Verdict Index route cannot yet deliver a diff (one in-service snapshot; a changelog needs two), and there are no exceptions for our own stall. Full evidence and the embed snippet: the badge page.

External record — kept in two buckets, never merged

• Non-Settled external payers: 19 calls / $0.445 across 4 wallets (one repeat payer at the $0.10 route).
• Settled scout/inspector traffic: 35 purchases / $2.625, kept separate — it is an inspector's test traffic, not organic demand.
• Seller seeds never count as traction. The recheck pack has had no organic stranger sale yet.
• No confirmed third-party loop runs on the fleet yet.

Loop shelf

• Recheck watch pack (StallRow POST /v1/recheck/pack): $1.00 = 250 change-checks on ONE endpoint, 90 days, $0.004/check. Balance reads are free; exhausted packs return 402, expired 410, and never fall back to per-call billing.
• x402-loop-kits: monitor / pre-trade / inspector templates (this repo). First live run, 2026-10-26 19:36 UTC: one inspector recheck of Token Verdict /v1/verdict spent one credit from a seeded pack (balance 250 → 249 read back afterward), result BASELINE, $0.00 cash.
• x402-guard: buyer-side middleware — PyPI x402-guard 0.1.1 and npm @quincygb/x402-guard 0.1.2 (published 2026-10-26 via Trusted Publisher workflows, registry-verified) can spend recheck pack credits for drift checks (config recheck_pack_id / recheckPackId), warning at 25 and 10 credits remaining.
• verdict-loop: the reference buyer — runs daily 08:00 America/Toronto under a $0.10/day cap with a public log. Labelled self-funded seller demo, not customer traffic.

New since 2026-10-26 evening

• Fleet Pass (StallRow POST /v1/fleet-pass): one $5.00 prepaid USD balance that spends on any paid route across all five services, debited at each route's listed price, 90 days. Send X-Fleet-Pass; balance reads free at POST /v1/fleet-pass/balance.
• Webhook Watch (StallRow POST /v1/watch/register): register an endpoint + webhook, funded by a recheck pack or a Fleet Pass — changes are pushed to your webhook as they are detected. First live registration 2026-10-26: BASELINE delivered to the test webhook (HTTP 200, payload verified at the receiver).
• Framework shelves: Token Verdict for ElizaOS, LlamaIndex, and the OpenAI Agents SDK — see frameworks/ in the loop-kits repo.
• Verdict Index (StallRow POST /v1/index/changes, $0.10): the CDP Bazaar catalog changelog — diffs of full-catalog snapshots (baseline 2026-10-26: 78,057 unique resources). Issue #0 is in the repo (verdict-index/ISSUE-0.md). In-service diffs start once two snapshots accrue; until then the route 502s and charges nothing.
• Orchestrate (StallRow POST /v1/orchestrate, $0.75): one combined report — Token Verdict + RouteVerdict + LockSmith + ListSmith via the fleet's internal paths. Live and 402-verified; its legs activate when the fleet-internal secret is set in production config (a dashboard step), so it is not seeded yet — a route that cannot deliver does not get listed.

Managed Loop — done-for-you monitoring

Don't want to run a loop yourself? We run it: change alerts pushed to your webhook plus a weekly report, on the same machinery this page reports on. Tiers: Watch $25/mo (1 endpoint, daily) · Row $79/mo (up to 5 endpoints, 6-hourly) · Fleet $199/mo (up to 25 endpoints, hourly) — pricing math, grounded in the fleet's own listed route prices, is in managed-loop/OFFER.md in the loop-kits repo. Client #0 is our own fleet; its demo report is published there (managed-loop/DEMO-CLIENT0.md). No external client is claimed — there isn't one yet.

Known limitations (stated, not hidden)

• No uptime figure is published, because none is measured. Liveness is probed by the daily integration watch; this page shows the check time behind every LIVE mark.
• No app-level rate limiting. A flood of well-formed forged payment headers would be forwarded to the facilitator for verification (2026-10-25 audit item M3, accepted risk, open decision). A facilitator-quota check on 2026-10-26 found that CDP's verify performs signature/nonce/balance and KYT screening itself, but no published per-seller verify quota was found in CDP's docs; the Railway-edge option requires dashboard access and has not been applied.
• Chunked request bodies without Content-Length terminate at the Railway edge; a DNS-rebinding TOCTOU residual exists in the SSRF guards (2026-10-25 audit).
• EntRoute is parked: its listing gate needs a DNS TXT record that cannot be set on *.up.railway.app, and no custom domain has been purchased.