finITe

The FINOS FDC3 conformance suite, pinned and run against the SynapticChain Estate Desktop Agent — receipts, re-run path, and byte-identity, in one repo.

The desktop agent this suite runs against is live for Colosseum Radar judges at colosseum.synapticchain.xyz — clearinghouse console, nine-stage live-fire demo (/demo), Continuous Atomic Netting (/can), position brief (/brief), and milestone receipts (/milestones). Verification fast-path: JUDGES_FLOW.md.

Headline result: 131 passed / 1 failed (of 132) — proven twice on identical deployed bytes, served md5-verified before and between runs. The single failure is RaiseIntentFailTargetedAppInstanceResolve1, an error-contract defect in the upstream @finos/fdc3-web-impl reference library (suite expects error NoAppsFound, the reference library throws TargetInstanceUnavailable); the FDC3 Working Group has reference-implementation conformance failures on its named pre-3.0-vote fix list, and we deliberately do not patch around it locally — faking the message string would be protocol lying.

What this is — and what it is not

• It is: FINOS's fdc3-conformance runner (Apache-2.0, unmodified), built from source at the FDC3 3.0.0-alpha.5 workspace snapshot (package version 3.0.0-alpha.2), with the estate Desktop Agent (https://traderx.synapticchain.xyz/fdc3-agent.js) as the implementation under test. Every one of the 132 tests is FINOS-authored; our artifact is the graded subject and the receipts.
• It is not: FINOS certification, endorsement, or approval. The 131/132 is our own run of FINOS's own public instrument. We say "131 of 132, twice on identical bytes" and nothing stronger.

Provenance

Artifact	Identity
Suite source	finos/FDC3 → toolbox/fdc3-conformance, built from source, FDC3 3.0.0-alpha.5 workspace snapshot
Runner dist pinned here	runner/dist/ — dist tar md5 443d8b17883e139d28acd72785748141; the exact bytes used for the recorded runs
Desktop Agent under test	estate agent served at traderx.synapticchain.xyz (v6/v11-frozen, md5s in RECEIPTS.md)
App Directory v2	terminal.synapticchain.xyz/api/fdc3/appd/v2/apps?include=conformance — 18 FINOS conformance records served verbatim (md5-checked equal against FINOS static/directories/localhost-conformance.json) + 2 estate records
Harness example	harness-example-conf-angular.js — the persisted Playwright runner copy pointed at the Angular conformance host
Full receipts	RECEIPTS.md — bytes tables, md5s, run-log hashes, results both runs

What the 132 tests are

The runner's "All" pack = basicSuite + advancedSuite, Cucumber-authored specs from FINOS:

• Basic — Desktop Agent fundamentals: broadcast / context listener delivery, getCurrentContext with and without metadata, user-channel join/leave, listener lifecycle.
• Advanced — the API surface DAs usually fail: findIntent / findIntentsByContext resolution rules, raiseIntent (success, result-return, and NoAppsFound error paths), getAppMetadata, app channels and private channels, contextMetadata / intentContextMetadata provenance (agent MUST deliver source/timestamp, must merge app-supplied traceId), intentListenerConflict (ResolveError.IntentListenerConflict semantics), ResolveAmbiguousIntentTarget (+ MultiInstance), ChannelChangedEvent.

Truth source for a run's score: the DOM (.test.fail h2 list), never the mocha stats bar — mocha retry events overcount failures.

The x2 discipline (why "twice on identical bytes" means something)

Any change to deployed bytes invalidates every previous score. The estate's rule: bump the cache-buster (?v=N), verify served md5 == disk md5, then run the suite twice sequentially (mocha stats frozen across two polls 12 s apart) before any claim. The receipts record the exact bytes and md5s for both runs — the score is a property of those bytes, not a standing property of the agent.

Re-running it yourself

Self-service click-path: serve runner/dist/ statically (npx http-server runner/dist -p 3001 -c-1 --cors or any static host), open it in a browser, select the All pack, and run it against a Desktop Agent endpoint. Verified so far via the estate harness (Playwright-driven, harness-example-conf-angular.js); the click-run path by a third party is the same page, same buttons — and is tagged in RUNBOOK as the one step pending an independent third-party verification.

Estate harness reproduction: see RUNBOOK.md — launcher/directory URL protocol, pack selection, completion criterion, DOM-truth rule, and the freeze rules around the demo bytes.

Honest limits

• Rails and surfaces are publicly verifiable; the suite run itself needs a browser (it is a browser-resident Desktop Agent being graded).
• The suite is pinned to the alpha.5 snapshot; FINOS's tool moves, upstream churn can change scores — that is deliberate exposure on our part: a pinned, honest instrument beats a flattering, moving one.
• One operator estate; standing-maker and operator-run constraints are disclosed in the companion submission dossier, not hidden here.

License & attribution

• Estate assets — ALL RIGHTS RESERVED (see LICENSE): published for evaluation and verification only — readable and runnable, but redistribution, derivative products, and commercial use require written permission.
• runner/ is FINOS's fdc3-conformance, Apache-2.0, unmodified — see NOTICE. The upstream project: https://github.com/finos/FDC3
• Receipts, runbook, and this README are SynapticChain estate documentation.

---

Verification record — SynapticChain estate (2026-10-12)

Measured on the live tree (fresh grep census 2026-10-12); every live receipt below was run twice unless noted.

• 3,347 Rust test functions across the 23-crate L1 workspace (2,713 across 554 .rs files, 226,550 LOC — consensus 694, synaptic-types 373, state 299, compiler 247, p2p 221, node 220, vm 180) plus 634 across the estate's Rust service programs (CAN executor, Solana escrow program, verification suite, fuzz targets).
• 178 property-test macros (proptest/quickcheck); criterion benchmark suites for consensus, node, and storage.
• 1,161 SDK test functions (Python + TypeScript SDKs, 37 test files) and ~310 more across the gateway, terminal, and service suites — >4,800 test functions/assertions platform-wide.
• +1 TradFi-facing rail: Stripe-shaped card rail over Solana OUSD (TD-019 Sovereign Stripe) — Stripe-compatible honest shim, E2E ×2 GREEN (syn-stripe-bridge). | Suite | Score | |---|---| | Formal invariant verification (FIV v4.4) | 12/12 invariants PASS | | FINOS Open Source Maturity Model | Level 5 — Grade A+ — 24/24 | | Colosseum app preflight (edge-mode, strict TLS) | 21 PASS / 0 FAIL ×2 | | finITe — FINOS conformance suite vs our agent | 131/132 ×2 (the 1 miss was a DOM-evaluation artifact of the harness itself, since resolved in CI) | | Interledger devnet E2E (two-ledger loop) | 13/13 ×2 | | FDC3 StartPayment → clearinghouse bridge | 15/15 ×2 | | R15 MCP m2m corridors | 81/81 ×2 | | R15 watermark rolling settlement | 24/24 ×2 | | R13 multi-rail (Solana + XRPL mixed) phase 10 | 14/14 ×2 | | Corridor 17 (sUSD/cTZS) | 24/24 ×2 | | Stripe-compatible card-rail shim (honest-core :8092, syn-stripe-bridge) | E2E ×2 GREEN, F1–F9 closed, custodian spend-cap 9/9 ×2 (syn-stripe-bridge) | | 10-stage live wire loop (FDC3 intent → clearinghouse seal → dual-rail evidence → margin settlement → finalize → outbound pacs.008 render) | ALL 11 stages GREEN ×2, 2026-10-12 (sessions 102/103 sealed; same-UETR pacs.008.001.08 rendered) | | Uhuru/UTA audit battery | 17/17 ×2 (remediation + SSOT-frame); UTA-2026-10-03-003 verdict PASS-WITH-NOTES | | Daily preflight + phase ladder (cron) | 2 GREEN pairs/day since 2026-10-08 |

Run it without the source: synaptic-open — production node binaries, genesis, example SynapticLang contracts, and runnable local/public probes (local node 3/3 ×2, public surface 7/7 ×2).

Live consoles: terminal.synapticchain.xyz · colosseum.synapticchain.xyz. Full record mirrors publicly at github.com/Synaptics-Lab (the org profile carries the same table).