Products

The White Report MCP

Free read-only crypto + AI news for agents: search articles, get full article text, live ticker, and a graded prediction scoreboard. Hosted at thewhitereport.com and listed in the official MCP Registry.

Platform: HTTPS (hosted) · Free public tier

Usage 30d: 1 calls · MEASURED · upd 2026-10-26

Eclipse Agent Wallet

The agent wallet where responsibility follows scope — per-transaction and daily caps, recipient allowlists, human approvals over a threshold, an instant kill switch, and an append-only audit log.

Platform: MCP over stdio · Open reference implementation · Sandbox only — simulated funds; production payment rails are not wired.

Usage: not collected — by design

CertBridge

Agent-native ACME certificate lifecycle — issuance, HTTP-01 challenge placement, download, and renewal watching for shared-hosting sites, without the dashboard paywall.

Platform: MCP over stdio · Bench-certified in Let's Encrypt staging; live production issuance proven on our own domain (eclipsebazaar.com, Oct 2026).

Usage: not collected — by design

HostBridge

The sanctioned path for agents to deploy websites to shared hosting: scope-bound deploys, snapshot-before-deploy, one-call byte-exact rollback, and a plain-language deploy receipt.

Platform: MCP over stdio · Certified against a local simulator docroot; live-host transports are declared, not yet live.

Usage: not collected — by design

Eclipse Connect

Hosted OAuth relay: the owner authorizes once in a browser, the relay holds the grant encrypted, and an agent uploads only owner-approved YouTube Shorts — never holding the channel's credentials.

Platform: HTTPS relay (hosted) · In live use for The White Report · Multi-user production scale is P2, not yet productized.

Usage: UNREACHABLE — relay audit log unreachable at snapshot time; no counts are shown rather than estimated.

Eclipse Bazaar

A marketplace for tested agent tools: agents discover products by the problem they solve, and every listing carries its Test Bench verdict — or states plainly that it has none. Discovery works today; purchasing is a later phase.

eclipsebazaar.com · Catalog: 33 products, statuses as labelled on this page

Usage: not collected — by design

Eclipse Herald

The agent-to-agent discovery and delegation layer: signed A2A Agent Cards for hireable services, a policy-gated gateway with certify-before-you-hire, AP2-shaped mandates with budget caps and expiry, and receipt-on-delivery in a hash-chained ledger. Foreign-agent output is ingested as quarantined data, never as instructions.

Platform: MCP over stdio · Test Bench 2026-10-25: 18/19 checks · local build; settlement in zero-value test credits

Usage: not collected — by design

Eclipse Beacon

The insurance watcher for a household: declarations pages become a provenance-carrying policy graph, an expiry radar signals renewals and deadlines before they happen, and a gap report names uncovered assets, duplicate coverage and premium spikes. It does not insure, bind, or advise.

Platform: MCP over stdio · Test Bench 2026-10-25: 16/17 checks · local build; read/watch only; synthetic fixtures

Usage: not collected — by design

Eclipse Cluster

The container gate: before an image is allowed anywhere near a runtime, Cluster verifies its cosign-style signature and DSSE provenance against a written policy and a local trust bundle — per-check PASS/FAIL, fail-closed on every error. v0.1 is the K0 verification gate only; orchestration is not built.

Platform: MCP over stdio · Test Bench 2026-10-25: 18/19 checks · local simulator; signed-image verification gate

Usage: not collected — by design

Eclipse Thread

On-prem bridges for the legacy software-engineering toolchain, plus the Evidence Weaver: Jira-shaped, SARIF, coverage and model exports join into one artifact graph with typed trace links and provenance, gap-first trace analysis, and deterministic, hash-chained evidence packages an auditor can verify independently.

Platform: MCP over stdio · Test Bench 2026-10-25: 13/14 checks · simulator build; fixtures/mocks only

Usage: not collected — by design

Eclipse Constellation

The governed, vendor-neutral VM fleet layer for agents: a declared-target registry (undeclared machines are invisible), wallet-transposed policy with quotas and image/network allowlists, a mechanical snapshot-before-mutation precondition, and a hash-chained ledger in which refusals are receipts too. v0.1 is observe-only.

Platform: MCP over stdio · Test Bench 2026-10-25: 15/16 checks · local build; observe-only, no mutation paths

Usage: not collected — by design

Eclipse Deep Field

The governed database layer for agents: table allowlists (unlisted tables are indistinguishable from nonexistent), read-only by default, an EXPLAIN cost gate that refuses full scans, PII masking before results leave the tool, and writes only inside scoped, expiring mandates via dry-run with before-image. Every query — blocked ones included — appends a hash-chained receipt.

Platform: MCP over stdio · Test Bench 2026-10-25: 16/17 checks · local build; SQLite full, Postgres live path not yet tested

Usage: not collected — by design

Eclipse Pantry

The unattended-retail module: reads a vending fleet's own DEX telemetry into provenance-carrying snapshots — per-selection inventory with staleness flags, sales summaries with the cash/cashless split, days-to-empty stock-out forecasts (labeled heuristics), and fault reads. Tier-0 read-only; a PAN-shaped string anywhere in an export rejects the file, and the digits are never stored or echoed.

Platform: MCP over stdio · Test Bench 2026-10-25: 14/15 checks · simulator build; read-only DEX telemetry

Usage: not collected — by design

Eclipse Marquee

The live-events agent, fan side first: normalized discovery that preserves section, row and seat semantics, face value and fees always itemized separately, and a Price Watcher whose ceilings compare only the all-in per-ticket price and may only tighten — loosening is rejected by name. v0.1 is read-only; no purchase or transfer surface exists.

Platform: MCP over stdio · Test Bench 2026-10-25: 15/16 checks · local stub feed shaped like Ticketmaster/SeatGeek responses

Usage: not collected — by design

Eclipse Dispatch

The newsletter layer above the platforms: one canonical, versioned, hash-bound draft syndicated as per-platform drafts to beehiiv and Ghost via sanctioned-API-shaped adapters, a paste pack (never a write) for read-only Substack, and unified analytics that carry each platform's own metric definition — differing definitions are flagged, never averaged. It never publishes without approval.

Platform: MCP over stdio · Test Bench 2026-10-25: 13/14 checks · local build; stub adapters, live wiring owner-gated

Usage: not collected — by design

Eclipse Social Core

One canonical, hash-bound content package behind every Eclipse social adapter: per-platform native variants with classified adaptation diffs (a rewrite that drops a canonical number is flagged substantive), hash-bound approvals that any later change invalidates, the house engagement caps enforced as policy, and per-account kill switches. It publishes nothing itself — it makes sure nothing unapproved moves.

Platform: MCP over stdio · Test Bench 2026-10-25: 15/16 checks · local build; governance core for the adapter family

Usage: not collected — by design

Eclipse Aperture

The Instagram adapter, formalized: one approved package becomes native Feed, Carousel, Story and Reel variants, with a preview pack rendering the exact native preview an approval binds. v0.1 codes the governed wrapper around the house's sanctioned connector route — no token path, session handling or posting code exists in this server, and its capability declaration says so.

Platform: MCP over stdio · Test Bench 2026-10-25: 11/12 checks · local build; formalization only, no live calls

Usage: not collected — by design

Eclipse Zephyr

The Bluesky adapter and the family's open-protocol reference implementation: real public reads over the AT Protocol public API — profile, actor search, author feeds, thread reads — normalized with per-item source endpoints, and every read receipted into a hash-chained ledger, failures included. No posting surface exists in v0.1.

Platform: MCP over stdio · Test Bench 2026-10-25: 12/13 checks · live public API reads; first CERTIFIED(live-read) in the catalog

Usage: not collected — by design

Eclipse Environment

The smart-environment module of Eclipse Procure: a Device Registry (unenrolled devices are invisible to every tool, even when the hub can see them) and a Safety Policy ladder — Tier 0 sense, Tier 1 benign, Tier 2 consequential, Tier 3 safety-critical — where owner overrides may only raise tiers, and a command earns a receipt only when the device itself confirms the effect.

Platform: MCP over stdio · Test Bench 2026-10-25: 14/15 checks · local HA-shaped stub; Tier 1 execution, Tier 2/3 deny-by-phase

Usage: not collected — by design

Eclipse Procure

The procurement pipeline for agents: intent → identity check → quote → Agent Wallet-shaped policy authorization → payment authorization → checkout → hash-chained receipt ledger → order tracking, fail-closed at every stage. A quote that cannot be fully priced can never be authorized, and a changed final total aborts the order before any payment record exists. v0.1 runs end to end in a simulator (fixture merchants, TEST rail); its modules — Zenith, Stage, Environment — are listed on this page; Vigil (medical guardian) remains gated and unbuilt.

Platform: MCP over stdio · Test Bench 2026-10-25: 14/15 checks · simulator end to end; no real merchants or money

Usage: not collected — by design

Eclipse Gig Core

One approved Task Brief for every task marketplace: versioned and hash-bound (any post-approval edit invalidates approval; exact street addresses are refused — ZIP-tier only), quotes normalized across lead, stated-budget, fixed-price and hourly models with face price and itemized fees forever separate, and a hire that runs only inside a live Procure / Agent Wallet mandate plus the owner's person-confirmation.

Platform: MCP over stdio · Test Bench 2026-10-25: 18/19 checks · local build; shared core of the gig adapter family

Usage: not collected — by design

Eclipse Compass

The Thumbtack adapter: an approved Task Brief becomes a native Thumbtack request — the platform category mapped explicitly (an unknown category is a named error, never a silent nearest-match), the category's own request form answered from the brief with every difference classified, unanswered fields surfaced as diffs rather than invented answers, and pro license claims shown verbatim with provenance, never vouched.

Platform: MCP over stdio · Test Bench 2026-10-25: 14/15 checks · Thumbtack-shaped fixtures; no live calls

Usage: not collected — by design

Eclipse Stage

The console bridge: deploy and playtest AI-built games on the owner's own Dev Mode console — device info, app records, metadata-only install simulation, controller input from an allowlisted vocabulary (validated as a whole before the first event is sent), frame capture, and power control — behind a Console Registry whose enrollment requires a dev-mode attestation flag. A retail-mode console is invisible to every tool, even if its portal answers.

Platform: MCP over stdio · Test Bench 2026-10-25: 15/16 checks · Device Portal-shaped simulator; real-console parity unverified

Usage: not collected — by design

Eclipse Zenith

The fitness module of Eclipse Procure: an FTMS bridge core (treadmill, bike, rower, heart rate) proven against hand-computed byte fixtures in both directions, an Equipment Registry, and a session state machine whose ending writes the workout record. Treadmill speed is Tier 3 behind a confirmation token — and the dead-man rule lives in the equipment layer: lapse the control heartbeat and the belt stops itself, receipted as a safety action.

Platform: MCP over stdio · Test Bench 2026-10-25: 16/17 checks · equipment simulator; real hardware/BLE unverified per model

Usage: not collected — by design

Eclipse Apply

Applies to jobs inside a human-granted written policy — target roles, salary floor, locations, allowlist + blocklist, daily cap, approval threshold, expiry, kill switch. Review-first: matches land as prepared applications, and tailored resumes re-order verbatim experience only, each with a machine-verifiable diff. v0.1 submissions are dry-run only — nothing is transmitted.

Platform: MCP over stdio · Test Bench 2026-10-25: 17/18 checks · sandbox environment; submissions dry-run only

Usage: not collected — by design

AgentReady

Can an AI agent actually buy from your store? Audits any public storefront for agent readiness — UCP profile, live MCP endpoints, structured product data — and returns a 0–100 score with observed evidence and top fixes.

Platform: MCP over stdio · Test Bench 2026-10-25: 13/14 checks · local build; public-surface audit only · Free local audit; hosted monitoring tier planned, not launched

Usage: not collected — by design

Florida Bid Intel

Florida bid intelligence with honest deadlines: MyFloridaMarketPlace search over dated snapshots, document-parsed due dates that outrank the portal's misleading grid dates, and a transparent fit score against a NAICS/commodity profile.

Platform: MCP over stdio · Test Bench 2026-10-25: 13/14 checks · local build; dated snapshots only — no live-pull path exists · Internal dogfood today; gaps the portal doesn't expose are reported as not exposed, never guessed

Usage: not collected — by design

Eclipse Bazaar MCP

The Bazaar's agent-facing discovery server: search products by capability, find solutions by problem statement, compare side by side, and look up any product's latest certification — including what was not tested.

Platform: MCP over stdio · Test Bench 2026-10-25: 15/16 checks · local build; standing card run 4, runs 1–3 FAILED retained (bench-suite defects) · Free discovery (CETech catalog today)

Usage: not collected — by design

Eclipse Muster

The shared core of the Eclipse customer family: a canonical Customer Graph (Person, Company, Deal, Ticket, Conversation, Activity) with provenance on every field and a Data Rights Ledger on every record. Identity merges are propose-only and exactly reversible; reply drafts cite record facts or mark them MISSING_FACT; instruction-shaped customer text is quarantined as data under Herald Security Law #1.

Platform: MCP over stdio · Test Bench 2026-10-25: 15/16 checks · Customer Graph core; local build, synthetic fixtures only

Usage: not collected — by design

Eclipse Haven

The governed Zendesk server, built over the platform's REST API v2 shape — the largest support base has no official MCP server. OAuth-first by construction: API-token auth is not implemented at all (Zendesk sunsets tokens 2027-04-30) and presenting one is refused by name on all 10 tools. Public replies send only as approved drafts; deletes, merges, bulk edits, and refunds are prohibited in v1.

Platform: MCP over stdio · Test Bench 2026-10-25: 13/14 checks · Zendesk-shaped fixture build; run-2 FAILED card retained, run 3 standing

Usage: not collected — by design

Eclipse Orbit

The governing proxy in front of HubSpot's official MCP server — the Eclipse WRAP posture. Vendor tools pass a pinned classification registry: unclassified tools are invisible, T3 tools are never surfaced, and a vendor tool whose tier rises is halted with a notice receipt until an owner clears it by name. T2 sends require approval tokens bound to the exact arguments.

Platform: MCP over stdio · Test Bench 2026-10-25: 16/17 checks · proven against a local HubSpot-shaped stub, both drift fixtures included

Usage: not collected — by design

Eclipse Voice Core

The shared core of the Eclipse voice & telephony family: the Consent Registry and Number Registry every telephony adapter obeys. Consent is provenance-bound — a record without provenance is void; revocation is one-way and effective at the next check everywhere, because adapters hold no consent copies. The send gate evaluates consent first, then an Agent Wallet-shaped mandate; an unconsented attempt writes a violation record and nothing else.

Platform: MCP over stdio · Test Bench 2026-10-25: 19/20 checks · Consent + Number Registry; local build, fixtures only — no carrier, no real numbers

Usage: not collected — by design

Eclipse Trunk

The governed Twilio adapter: raw carrier operations are never exposed — log reads with numbers masked beyond last-4, simulated sends, and recording policy, every send gated through Voice Core (consent, then mandate, then number sendability). Simulated human calls require a named human speaker; there is no AI voice in this adapter.

Platform: MCP over stdio · Test Bench 2026-10-25: 16/17 checks · Twilio fixture adapter; V0 reaches no carrier — nothing sent, nothing provisioned

Usage: not collected — by design

Eclipse Connect MCP

The MCP face of Eclipse Connect, the hosted OAuth bridge: agents upload, publish, check grant status, and revoke through a revocable agent key — never raw OAuth. Going public requires an explicit human approval flag, enforced in the facade before any network call and again at the relay. A distinct surface from the relay itself, and benched as one: the relay holds its own CERTIFIED(dogfood).

Platform: MCP over stdio · Test Bench 2026-10-25: 14/15 checks · facade over the live single-tenant relay (Testing-mode OAuth)

Usage: not collected — by design

Eclipse Envoy

The signed front door for agent-to-agent business: one owner-approved identity source becomes a Herald-signed corporate Agent Card; counterparties are vetted into risk tiers R0–R3 (typosquats get their own record and never touch the honest org’s history), disclosures render per tier with an owner-entered NDA gate, and engagements bind only by the three-key rule — owner approval + Wallet mandate + Herald delegation. No tool on either surface signs, commits, or moves funds.

Platform: MCP over stdio · Test Bench 2026-10-25: 17/18 checks · corporate agent interface; local build, offline fixtures only — public use of the name is gated on a trademark screen

Usage: not collected — by design

Eclipse Fathom

DeFi, measured honestly: a signed, versioned Protocol Registry (Uniswap/Aave/Compound/Lido on Ethereum/Arbitrum/Base; audit status as dated data, never scraped into trust), dual-source position reads that render DIVERGENT instead of averaging, depeg watch, lookalike-address checks, and the flagship allowance audit that ranks unlimited ERC-20 + Permit2 grants by value at risk. Reads only — no execution, draft, or signer exists, and key-shaped input is refused by name.

Platform: MCP over stdio · Test Bench 2026-10-25: 16/17 checks · DeFi risk core; fixture transport — no live chain queried; live RPC reads unverified

Usage: not collected — by design

Eclipse Codex

Banking regulation as data: a versioned Regulation Library (records never overwritten, as-of resolution, NPRM text never resolves as law), an Obligations Mapper whose merge guard keeps the CDD Rule and CTA BOI duties separate, a Deadline Engine that never lets a guidance cadence render as a breach, local sanctions screening with no auto-clear, and evidence packs with a mandatory gap section — suppression is refused and receipted. Not a bank, not counsel, not a compliance department: it files nothing and determines nothing.

Platform: MCP over stdio · Test Bench 2026-10-25: 18/19 checks · banking-regulations core; synthetic fixture institution only — R1 design partner + counsel review ahead

Usage: not collected — by design

Eclipse Ledger

Betting records and responsible-gambling controls. Not a sportsbook. No bets are placed here. Limits fall instantly but rise only after the stricter of the suite floor or the state’s delay plus a fresh confirmation; self-exclusion is absolute, honored identically from any source, with bookkeeping preserved; the jurisdiction matrix treats sportsbook, DFS, and pick’em variants as distinct legal objects; and P&L is honest cash math — a bonus bet’s stake is never counted as money. No placement tool, no account connection, no credential accepted.

Platform: MCP over stdio · Test Bench 2026-10-25: 21/22 checks · RG core; fixtures only — live users wait on the spec’s counsel gate

Usage: not collected — by design

Eclipse Till

Bazaar checkout for agents, as its own server — the discovery Bazaar MCP’s certified no-purchasing surface is untouched. Quotes bind catalog + pricing version + terms and state the all-in total; the mandate gate checks cap against the all-in, not the sticker; over-threshold purchases park for the buyer’s own approval with no pressure verbs anywhere; entitlements are Herald-signed and buyer-bound; and a missed delivery window fires refund + revocation automatically, with no seller override. B0 settles sandbox test credits only — no real rail code exists in the build.

Platform: MCP over stdio · Test Bench 2026-10-25: 23/24 checks · purchase core; test credits only — Stripe/x402 rails are B1, owner-gated

Usage: not collected — by design

Eclipse Book

The Markets core: one canonical Instrument / Book / Quote / Order / Position / Resolution model across exchanges and prediction markets. Criteria are stored verbatim and hash-pinned — a venue edit is a CRITERIA_CHANGED event and positions keep their criteria as of entry; quotes carry all-in cost from versioned fee schedules or degrade to FEE_UNVERIFIED; resolutions verify only against the criteria’s own named source, with DISPUTED first-class; and no output ever carries a probability field — the guard is code, not policy. Market data + venue access layer — not an exchange.

Platform: MCP over stdio · Test Bench 2026-10-25: 16/17 checks · markets core; data only, fixtures behind a transport seam — no accounts, no keys, no order submission

Usage: not collected — by design

Eclipse Forecast

The Polymarket adapter: Gamma/CLOB-shaped data normalized into the Book core’s canonical instruments and books, token/outcome mapping pinned from the venue’s own arrays, criteria verbatim + hash-pinned, fee-truth quotes, and UMA-shaped resolutions with DISPUTED as a first-class state. Every output carries money_class: REAL. The US trading surface is absent by design — no order, trade, account, or API-key tool exists, and circumvention-shaped names are refused by name.

Platform: MCP over stdio · Test Bench 2026-10-25: 13/14 checks · prediction-market adapter; fixtures behind a transport seam — live venue reads are a later verification step

Usage: not collected — by design

Eclipse Strike

The Coinbase public-data adapter: products and level-2 books normalized through the Book core (the same builders Forecast uses — cross-adapter parity is bench-proven), native statuses the core cannot map preserved verbatim as UNMAPPED, and quotes carrying all-in cost or FEE_UNVERIFIED. Coinbase ships its own official MCPs for accounts and wallets; Strike re-implements none of them — no account tools exist in this build.

Platform: MCP over stdio · Test Bench 2026-10-25: 13/14 checks · exchange data adapter; fixtures only — account paths are an M1+ governed wrap

Usage: not collected — by design

Eclipse Corona

Sealed render-proof for deployed websites: renders the served artifact across a pinned engine × viewport matrix, runs a closed vocabulary of mechanical probes at each cell, and seals measurements + screenshot hashes into a signed pack bound to the deployed artifact's hashes — recomputed by anyone. The publish gate may report 'verified live' only while a PASS served pack's hashes equal the deployed manifest. A pack proves identity and mechanical soundness of the pinned bytes only — never that an asset is semantically the right asset; semantic approval is the separate human act on the contact sheet. Name CONDITIONAL — trademark screen pending.

Platform: MCP over stdio (+ Gate CLI + Sheet CLI) · Test Bench 2026-10-25: 25/27 checks · local verifier only, P0 ceiling — no hosted/live scope

Usage: not collected — by design

Eclipse Verbatim

Title 26 as a sealed object: the OLRC-published text (release point Public Law 119-124) is sealed at ingest under the published canonicalization verbatim-canon/1.1 — per-section hashes for all 2,277 sections, dual Merkle roots, a signed manifest — and cite-check renders a byte-for-byte verdict (MATCH / DIVERGES with the first divergence localized / NOT_IN_CORPUS) against the sealed text; the bench's hostile battery produced 0 false MATCH. An independent implementation written from the spec alone reproduces the seal exactly, 2,277/2,277 in both the statutory and editorial layers; OLRC editorial notes are sealed separately and never served as statute. No advice surface — no tool interprets, explains, applies, or calculates anything; Title 26 statute only, no CFR; a dated snapshot, never described as current law. Signing uses labelled Herald fixture keys only. Name CONDITIONAL — trademark screen pending.

Platform: MCP over stdio · Test Bench 2026-10-26: 5/5 scoped re-bench scopes PASS · snapshot usc26-119-124-de68dcc8b3db · local scope only

Usage: not collected — by design

Eclipse Roll

Filing prep + proof for business & government filings: a filing is prepared, approved, and evidenced as a Herald-signed delta between the registry-of-record's own current state — read back through its sanctioned public surface at prep time — and the exact artifact to be submitted; after the human submits via the registry's own channel, Proof re-reads the registry and seals proof-of-filing only if the public record changed exactly as approved. No submission verb exists on any surface, in any name — prep-and-seal only, the human always submits; advice, selection, interpretation, and clearance requests refuse at the tool layer by name (ADVICE_CLASS_REFUSED). Certified at FIXTURE scope: every Gauge adapter is a labelled fixture and no live registry read has been exercised yet; dogfood subjects are CETech's own Florida annual report, SAM.gov renewal watch, and VetCert entry, as fixtures. First bench FAILED (23/34) and is retained on the record; the fix round closed D1–D10 and the scoped re-bench passed 17/17. Name CONDITIONAL — trademark screen pending.

Platform: MCP over stdio · Test Bench 2026-10-26: 17/17 scoped re-bench checks · fixture scope · counsel gate stands before any live phase

Usage: not collected — by design

Eclipse Verbatim — Executive Orders

Executive Orders as sealed objects: Federal Register full-text XML sealed per section under canon-eo/1.0 into snapshot eo-20261022-26818c895dee — 1,315 orders, per-section hashes, a signed manifest — plus a disposition graph extracted from the orders' own sealed sections: 10,667 edges, each pinned to the hash of the asserting section and typed by provenance (TEXT / FR_NOTE / NARA_TABLE). Where the consulted sources disagree, the disagreement is sealed as a DIVERGENCE record — 823, OPEN, never resolved by the product. Coverage is measured, not assumed: sealed text runs from 2000 (the Federal Register full-text XML floor); 258 orders are EXCLUDED_NAMED, and pre-2000 orders ride in a labelled unsealed metadata+edges layer reaching back to 1931. No tool characterizes legal effect — no status field exists anywhere in the product; a disposition answer is an edge set + as_of + a coverage window + the standing completeness sentence. The snapshot seal is a labelled FIXTURE seal. Name CONDITIONAL — inherits the pending Eclipse Verbatim trademark screen.

Platform: MCP over stdio · Test Bench 2026-10-26: CERTIFIED(local) scoped re-bench #3 · snapshot eo-20261022-26818c895dee · local scope only

Usage: not collected — by design

Eclipse Flood Witness

Flood evidence over sealed snapshots of FEMA's public National Flood Hazard Layer: NFHL county extracts are sealed per partition under flood-canon/1.0, and geometry determinations (FOOTPRINT / STRUCTURE_POINT / PARCEL — an address is refused as a subject) are minted as signed, independently recomputable receipts; the standalone verifier ships in the tree. Two Florida partitions are sealed — Leon (snapshot nfhl-fl-leon-20260611-bec93b592c31; 12,099 features) and Monroe (nfhl-fl-monroe-20261022-4776f79b6d7c; 9,664 features); the operative index enumerates all 67 Florida counties and names the other 65 NOT_SEALED — Miami-Dade is NOT_SEALED (pending). Evidence-grade only, by construction: never a determination of record, never an SFHDF, never lender-compliant, never an insurance answer, and never a statement that a property is "safe." Flag-1 qualifier: the LOMC override branch is fixture-proven only — the sealed corpus holds zero LOMA/LOMR-F records and the branch has never fired on real sealed data. One LOW (FW-D8) stands as a printed pre-second-vintage condition. All seals are FIXTURE_SIGNED and labelled as such; nothing here states or implies FEMA endorsement — FEMA/NFHL name the public data source only. Name CONDITIONAL — trademark screen pending.

Platform: MCP over stdio · Test Bench 2026-10-26: CERTIFIED(local) scoped re-bench · 7/7 fix-round dispositions verified · Leon + Monroe sealed scope only

Usage: not collected — by design

Eclipse Recall Witness

Recall campaigns, witnessed: NHTSA serves only the current state of a campaign — amended filings replace records in place and no vintage parameter exists — so this product seals the recalls flat file nightly into a hash-chained anchor and seals every campaign's full payload as an immutable vintage under recall-canon/1.0; a campaign that disappears seals as a tombstone, never a silent drop. Drift between vintages seals as dated events across the closed recall-events/1.1 grammar (POPULATION_EXPANSION · REMEDY_CHANGE · ESCALATION_DO_NOT_DRIVE · ESCALATION_PARK_OUTSIDE · CORRECTION · CLOSURE) — an escalation is never absorbed into an expansion. Posture-at-date derives exclusively from sealed vintages (NOT_SEALED before the first vintage, with the campaign's own first-vintage date), and mandate-bound check receipts print scope_class: CAMPAIGN-LEVEL with the three standing sentences byte-exact — a specific VIN is never called “open” and per-VIN unrepaired status is never determined (the consumer VIN lookup is never automated; no such code path exists). CERTIFIED at FIXTURE scope: the nightly spine is fixture-proven and no live nightly seal exists; live cross-check and one live check issuance are proven against api.nhtsa.gov (24V809 seals cross_check OK, its DO_NOT_DRIVE tag agreeing with parkIt: true; rcpt-00001 for a Porsche 911 MY2024 recomputes MATCH) — and the spine was NOT re-routed to the API while www-odi.nhtsa.dot.gov is unreachable, because the flat file is the canon source; LIVE scope waits on the first real nightly ingest and live-key cutover. Churn is MEASURING — counters seal from night one, no rates before a day-90 segment seals. Three LOW conditions stand printed on the bench card, all riding the recall-events/1.2 / next-seal touch: RW-D7 (narrative “parked outside” text can tag a false PARK_OUTSIDE — no live-stock instance), RW-D8 (genesis links not yet asserted in the chain walk), RW-D9 (operate/use paraphrase families untagged — no escalation-completeness claim beyond the live-stock-proven phrase families). All seals are under FIXTURE ECDSA keys, labelled on every artifact. Pricing is not yet set — no price points exist in the price book for this product; nothing is priced or activated. Name CONDITIONAL — trademark screen pending.

Platform: MCP over stdio · Test Bench 2026-10-26: CERTIFIED(local, FIXTURE scope) scoped re-bench · 127/136 checks · RW-D1..D6 FIXED verified · live cross-check + check issuance proven · nightly spine fixture scope only

Usage: not collected — by design

BidMatrix

Drop in a solicitation, get back the compliance matrix: Section L/M extraction, requirements checklist, deadlines, and set-aside summary.

Usage: not collected — by design

Section 508 Evidence

WCAG + Section 508 audits that produce the deliverable federal buyers actually require: ACR/VPAT-ready evidence reports.

Usage: not collected — by design

FedScope

Federal + state opportunity search with fit and newcomer scoring and plain-language summaries — not a raw API wrapper.

Usage: not collected — by design

MCP Health Check

Scan an MCP server or agent configuration for security posture — auth, SSRF exposure, static keys — with a scored report and fixes. Will dogfood on our own servers first.

Usage: not collected — by design

Legacy Recon

Point an agent at a codebase: architecture map, dependency / end-of-life risk list, and inputs for a modernization estimate.

Usage: not collected — by design

Eclipse Mail

A real business email identity for an AI agent — governed, review-first sending with scoped grants, plain-language receipts, a kill switch, and an audit trail. (Working name, not locked.)

Usage: not collected — by design