Download Free CE

SynapCores AML — The Open-Core Anti-Money Laundering Platform

Transaction monitoring with semantic alert dedup. UBO graph traversal in 100ms. Autonomous SAR drafting grounded in similar historical filings. Immutable audit log your FFIEC / OCC / FinCEN examiner can query directly. Managed cloud, or one docker compose on your infrastructure. Per analyst seat — no per-customer-screened tax.

Join the waitlistStar on GitHubEnterprise — request pricing

The pain we built this for

If you run a BSA/AML or financial-crime program, you have lost this conversation a hundred times: “Our transaction- monitoring system fires 50,000 alerts a month. 99% are false positives. Our analysts drown. Our SAR-narrative quality is inconsistent. Our examiner wants evidence, and we are still pulling it manually three weeks before the visit.”

  • Your TM analysts close 99% of alerts as false positive. Most are near-duplicates of last week's noise — nobody has time to dedupe semantically.
  • UBO traversal is a manual exercise across OpenCorporates + a national registry + somebody's Excel. By the time you reach beneficial-owner layer 5, the case is two weeks old.
  • SAR-narrative quality is analyst-dependent. The same fact pattern produces a great narrative from your senior, a mediocre one from your junior — and you have no consistency for the regulator.
  • Your examiner shows up. You spend three weeks pulling evidence — alert history, screening rationale, decision audit — into PDFs. There is no defensible chain-of-custody proof.
  • You evaluated ComplyAdvantage. The per-customer-screened pricing has a punishing growth curve. Hummingbird gives you a case manager but not the brain. Quantexa is too heavy and too expensive.

Six capabilities. One self-hosted stack.

Most AML platforms are a rules engine bolted onto a case manager. SynapCores AML is a Node.js application backed by SynapCores — our AI-native database with graph, vectors, immutable audit, SQL, and agentic loop in one engine. The category-defining capabilities that follow are not add-ons. They are native to the data tier the app sits on.

Property graph for UBO, structuring & mule networks

ComplyAdvantage and Hummingbird have workflow engines but no graph. SynapCores has Cypher built into the same database — a single MATCH query traces the ultimate-beneficial-owner chain up through 8 layers of holding companies, finds the mule cluster a flagged transaction sits in, and detects structuring patterns that batch-mode systems miss for days.

AI agents for triage, enrichment, and SAR drafting

Five autonomous agent personas — tm-triage, kyc-enricher, sanctions-investigator, sar-drafter, evidence-collector — run as SQL functions inside the database. The sar-drafter pulls similar historical SAR narratives via vector cosine, walks the UBO graph, and produces a regulator-templated 5W narrative the analyst edits and files.

Semantic alert dedup on transaction monitoring

TM analysts drown in alerts — 95-99% are false positives, most are near-duplicates of last week's noise. SynapCores embeds every alert as a 384-dim vector and cosine-matches against the rolling 90-day window. The duplicates auto-close with a one-line decision rationale; the analyst sees the actually-novel patterns, already enriched.

Cryptographically immutable audit log

Every alert, every screening hit, every agent action, every analyst decision, every SAR filing, every examiner query — written to an append-only table with cryptographic chaining. Run VERIFY_CHAIN() and prove the audit trail has not been tampered with. The surface that wins FFIEC, OCC, NYDFS Part 504, FCA SMCR, AUSTRAC, and FINTRAC examiner conversations.

MCP portal for examiners and internal audit

Mint a scoped, read-only, time-bound MCP token. The examiner or internal auditor pastes it into Claude or Cursor and queries the case file, audit log, and evidence packs in natural language. Every query they run is itself audit-logged. The screen that turns a 6-week examination cycle into a token and a chat.

Open core. Managed or self-hosted. No per-screen tax.

ComplyAdvantage prices per customer screened — the more customers you have, the more you pay forever. SynapCores AML charges per analyst seat, capped. The application source is public on GitHub. Choose managed cloud (we operate it) or self-host via one `docker compose up` — your data, your call, no devops project.

How it works

Three architectural pieces, all running on the same database connection.

1. The UBO + counterparty graph

Every customer, every account, every counterparty, every legal entity, every UBO, every transaction is a graph node. Every relationship (OWNS, CONTROLS, SIGNS_FOR, SENT_TO, DIRECTOR_OF) is an edge. One Cypher query walks 8 layers of holding companies to surface the ultimate beneficial owner, finds the mule cluster a flagged account sits in, and detects structuring patterns the batch-mode legacy systems miss for days.

Cypher in the same engine as SQL. Real-time, not nightly batch.

-- "Trace ultimate beneficial owners + structuring"
CYPHER
MATCH (tx:Transaction {id: $tx})-[:FROM]->(c:Customer)
MATCH ubo_path = (c)-[:OWNS|CONTROLS|DIRECTOR_OF*1..8]->(ubo:LegalEntity)
  WHERE ubo.is_natural_person = true
MATCH structuring = (c)-[:SENT|RECEIVED*1..3]-(peer:Account)
  WHERE peer.beneficial_owner = ubo.id
    AND tx.ts >  $window_start
RETURN c, ubo, ubo_path, peer, structuring
-- 100ms across millions of transactions.

2. Five autonomous agent personas

tm-triage handles every incoming TM alert — dedup via embedding cosine, run screening, score, escalate or auto-close with rationale. kyc-enricher does risk-based CDD/EDD on customer onboarding. sanctions-investigator resolves screening hits (true/false/inconclusive) with explainable evidence. sar-drafter retrieves similar historical SAR narratives via vector cosine, walks the UBO graph, and produces a regulator- templated 5W narrative. evidence-collector assembles tamper-evident examiner packs.

Each persona is a recipe in the public GitHub repo. Compliance officers customise their own.

-- Trigger autonomous triage on every new alert
SELECT AGENT_RUN('tm-triage', alert_id) AS verdict
FROM tm_alerts
WHERE status = 'new';

-- The agent's tool registry includes:
--   execute_query, vector_search, ubo_walk,
--   sanctions_screen, pep_screen, adverse_media,
--   draft_sar, auto_close_with_rationale,
--   escalate_to_human, ...
--
-- Each tool call is one row in the immutable audit
-- log. SAR-filing tools always pause for human
-- approval — there is no autonomous filing.

3. Immutable audit + MCP for examiners

Every alert, every screening hit, every agent action, every analyst override, every SAR filing, every examiner query — written to an IMMUTABLE table with cryptographic chaining. Run VERIFY_CHAIN() to prove the trail has not been tampered with. Mint a scoped, read-only, time-bound MCP token for your FFIEC / OCC / NYDFS / FCA examiner — they paste it into Claude or Cursor and query the case file directly. Every examiner query is itself audit-logged.

This is the screen that wins your next examination cycle.

-- The immutable audit table
CREATE IMMUTABLE TABLE audit_log (
  event_id    INTEGER PRIMARY KEY,
  ts          TIMESTAMP NOT NULL,
  actor       TEXT NOT NULL,
  actor_type  TEXT NOT NULL,
  -- 'analyst' | 'agent' | 'system' | 'examiner'
  action      TEXT NOT NULL,
  case_id     TEXT,
  sar_id      TEXT,
  payload     JSON,
  request_id  TEXT
);

SELECT VERIFY_CHAIN('audit_log') AS chain_ok;
-- → { chain_ok: true, last_event_hash: "0x…" }

SynapCores AML vs ComplyAdvantage vs Hummingbird vs Unit21

The honest comparison. We ship the capabilities the incumbents bolt on, and we charge for the things the incumbents charge for the wrong way.

FeatureSynapCores AMLComplyAdvantageHummingbirdUnit21
Open core (source on GitHub)YesNoNoNo
Managed cloud OR self-hosted (docker compose)YesSaaS onlySaaS onlySaaS only
Property graph for UBO + mule + structuringYesNoNoLimited
Native vector search (similar-SAR retrieval)YesNoNoNo
Autonomous AI agents (triage + SAR drafting)YesNoWorkflow onlyWorkflow only
Cryptographically immutable audit logYesNoNoNo
MCP portal for examinersYesNoNoNo
Pricing modelPer analyst seatPer customer screenedPer workflow / seatPer monitored entity
Data residency controlYour infrastructureSaaS-lockedSaaS-lockedSaaS-locked

Feature claims for third-party products reflect public documentation and customer commentary as of late 2026. We will correct anything inaccurate — open a PR on the homepage repo.

Who this is for

AML is bought and run by the financial-crime team — the BSA Officer, the MLRO, the head of financial crime. The primary buyer is process-rigorous and technical-enough to value an audit trail. For the CCO/CRO who signs the cheque, the managed-cloud tier means nothing to install or operate.

Primary buyer
BSA Officer / MLRO / Head of Financial Crime

Runs the AML program. Owns the budget. Has been burned by examiner findings before — wants an audit trail they can defend. Comfortable with technology when it makes the program defensible. Open-core is a pro signal (auditable code, no black-box decisioning).

Economic buyer
CCO / CRO / Head of Compliance

Signs the cheque and the regulatory attestation. Cares about examiner-defensibility, the per-customer-screened cost curve, and not getting paged at 2am. Uses Enterprise managed cloud so there is nothing to install. Open-core means board-defensible (no vendor lock-in, supply-chain transparent).

High-leverage
Neobank / PSP / MSB / Fintech CTO

Fast-growing. ComplyAdvantage per-customer-screened math is going to break the unit economics at 1M+ customers. Wants modern infrastructure (graph + vectors + agents) without bolting together five vendors. Self-host or managed-cloud — both work.

Quickstart — 30 seconds

The stack is one docker compose up away — no devops project. Point your core banking system or payment rail at /v1/aml/transactions and the triage agent goes to work. Prefer not to host anything? Enterprise includes managed cloud.

# 1. Clone the AML stack (AML app + SynapCores)
git clone https://github.com/SynapCores/synapcores-aml
cd synapcores-aml

# 2. Configure your screening-list keys + jurisdiction
cp .env.example .env
$EDITOR .env
#   → SCREEN_OFAC_FEED_URL=...
#   → SCREEN_PEP_PROVIDER=dowjones|refinitiv|comply...
#   → SAR_JURISDICTION=us-fincen|uk-nca|au-austrac|ca-fintrac

# 3. Bring up the whole stack — ~30 seconds
docker compose up -d
#   → synapcores (data tier)    on :8080
#   → synapcores-aml (Node app) on :3001

# 4. Get the admin password from the first-boot log
docker compose logs synapcores 2>&1 | grep "password:"

# 5. Point your core / payment rail at /v1/aml/transactions
#    Webhook, Kafka topic, or batch — your choice.
#
# Prefer not to host anything? Enterprise includes a
# fully managed cloud deployment — same stack, we run
# it for you.

Pricing

Open core: the application source is public on GitHub. The wedge against ComplyAdvantage is a per-analyst-seat model with no per-customer-screened tax — your AML cost stops scaling with your customer base. Self-host with one docker compose up, or let us run it for you in Enterprise.

Open Core

$0

Self-hosted via docker compose. Yours forever.

  • Full AML application source (Node.js / TypeScript)
  • SynapCores Community Edition as the data tier
  • One `docker compose up` brings the whole stack up in ~30s
  • All five agent personas (TM triage, KYC, sanctions, SAR draft, evidence)
  • Transaction-monitoring rule library (structuring, round-number, peer, velocity, …)
  • Sanctions/PEP/adverse-media screening adapters (BYO feed key)
  • UBO graph + Cypher traversal
  • Cryptographically immutable audit log + VERIFY_CHAIN()
  • Case-management UI + investigation graph explorer
  • GitHub + Discord community support
github.com/SynapCores/synapcores-aml
Enterprise

Custom

Per-analyst seat. No per-customer-screened tax.

Everything in Open Core, plus:

  • Fully managed cloud — we operate the stack, you log in (no devops required)
  • Assisted on-prem deployment if you prefer to host it yourself
  • 24/7 support + on-call escalation + dedicated solution engineer
  • Certified screening-list integrations (Dow Jones, Refinitiv, ComplyAdvantage data feeds)
  • Certified jurisdiction-specific SAR templates (US FinCEN, UK NCA, AU AUSTRAC, CA FINTRAC, EU GoAML)
  • Examiner portal (scoped MCP tokens, session recording, evidence pack builder)
  • Compliance certifications: SOC 2 Type II, ISO 27001, FedRAMP (roadmap)
  • Multi-tenant isolation for MSP and acquirer scenarios
  • Custom integration development (we wire your core / payment rail)
  • SLA-backed uptime + incident response from us
Request Enterprise pricing

Frequently asked

What is SynapCores AML?+

SynapCores AML is an open-core anti-money-laundering platform. The application is a Node.js/TypeScript service; the data tier is SynapCores, our AI-native database. It runs transaction monitoring, ultimate-beneficial-owner graph traversal, sanctions/PEP/adverse-media screening, autonomous SAR-draft agents, and a cryptographically immutable audit log for examiner defense. Deploys as a docker-compose stack you self-host, or as managed cloud we operate for you.

How is SynapCores AML different from ComplyAdvantage or Hummingbird?+

ComplyAdvantage is a closed-source SaaS data + workflow service priced per customer screened. Hummingbird is closed-source case management. SynapCores AML is open-core (the application source is public on GitHub), available as managed cloud or self-hosted, and priced per analyst seat — no per-customer-screened tax, no per-decision tax. It also ships with a property graph for UBO traversal, vector search for similar-SAR-narrative retrieval, and an immutable audit log built into the same database — capabilities you would otherwise stitch together from Neo4j, Pinecone, and a custom audit service.

Will my examiner accept evidence from SynapCores AML?+

Yes. Every alert, every screening hit, every analyst decision, every agent action writes to a cryptographically chained immutable table. Run VERIFY_CHAIN() to prove the audit trail has not been tampered with. Generate examiner-ready evidence packs in one query. For FFIEC, OCC, FinCEN, NYDFS Part 504, and FCA SMCR engagements, you can also mint a scoped read-only MCP token the examiner pastes into Claude or Cursor and queries the case file directly — every examiner query is itself audit-logged.

What does "open-core" mean — do I need a devops team to run it?+

No devops team required. The AML application source is public on GitHub — every transaction-monitoring rule, every screening adapter, every SAR template, every agent persona is reviewable code under a permissive license. To run it: one `docker compose up` brings up the stack (AML app + SynapCores) on your infrastructure in about 30 seconds. If you would rather not host anything, Enterprise includes managed cloud — we operate it, you log in. Open-core means the code is honest; how you consume it is your choice.

Does SynapCores AML replace my core banking system?+

No. SynapCores AML sits beside your core banking, payments, or ledger system. It ingests transactions via webhook or streaming (Kafka, Kinesis, RabbitMQ, direct DB-CDC), screens against sanctions/PEP/adverse-media lists, runs transaction-monitoring rules and behavioral models, walks the UBO graph, drafts SARs, and writes back to your case-management or core system via API. Your core system keeps source-of-truth ledger; we add the financial-crime brain on top.

What data sources does SynapCores AML ingest from?+

Core banking and ledger systems (FIS, Fiserv, Jack Henry, Mambu, Thought Machine, custom), payment rails (FedWire, ACH/NACHA, SWIFT MT/MX, ISO 20022, RTP, FedNow, SEPA, Faster Payments, card networks), KYC providers (Onfido, Persona, Alloy, Trulioo, Sumsub, Jumio), sanctions/PEP/adverse-media feeds (OFAC, EU, UN, HM Treasury, Dow Jones, Refinitiv World-Check, ComplyAdvantage, LexisNexis), corporate registries (OpenCorporates, OpenSanctions, national registries), and transaction-monitoring engines (Actimize, SAS, Verafin, Oracle Mantas) for migration.

How does the SAR-drafting agent work?+

When an alert is escalated to SAR, the sar-drafter agent retrieves the most similar historical SARs by vector cosine on narrative embedding, walks the UBO graph to surface ownership and relationships, pulls the transaction timeline, applies your jurisdiction's narrative template (FinCEN SAR for US, JMLSG/NCA for UK, AUSTRAC SMR for AU, FINTRAC STR for CA), drafts the 5W narrative grounded in the case evidence, and surfaces it for analyst review. The analyst edits, approves, and one click files the SAR via the regulator API. Every step is audit-logged.

Where is my customer and transaction data stored?+

Wherever you decide. Self-hosted: everything — customers, transactions, screening hits, case files, SAR narratives, audit logs — lives in your infrastructure inside the SynapCores RocksDB-backed engine. No customer data leaves your network. Managed cloud: dedicated single-tenant SynapCores instance in your chosen region, ours to operate, yours to query and export. Embedding generation and LLM inference can run locally on the bundled native model or be routed to your chosen provider — your choice, your contract, your data residency.

What does pricing look like?+

Open-core: no license fee for the self-hosted application. Enterprise is per-analyst-seat — no per-customer-screened tax, no per-transaction-monitored tax, no per-SAR-filed tax. Enterprise covers managed cloud hosting, 24/7 support and on-call escalation, the certified screening-list integrations, the certified jurisdiction-specific SAR template library, the examiner-portal license, compliance certifications (SOC 2 Type II, ISO 27001, FedRAMP roadmap), and dedicated solution engineering.

When will SynapCores AML be generally available?+

The open-core repository is live at github.com/SynapCores/synapcores-aml. General availability of the full enterprise tier is targeted for Q1 2027. Join the waitlist to lock in design-partner status, get architecture-review access, and shape the screening + SAR-template libraries.

Get on the SynapCores AML waitlist

Design-partner slots are limited. Waitlist members get:

  • First access to the public GitHub repo as it lands
  • A scoped architecture-review call with the engineering team
  • Input on the SAR-template library — your jurisdiction gets prioritised
  • Lock-in on Enterprise pricing for the first twelve months
  • Direct line to engineering for screening-feed and core-banking integration requests

Free weekly course

Learn machine learning, hands-on

Join the newsletter for a free weekly Machine Learning course you run yourself in Community Edition — plus new-release notes and support tips. No spam, unsubscribe anytime.

Free forever. We'll never share your email.

One stack. The whole financial-crime function.

Graph for UBO. Vectors for similar SAR narratives. Agents for autonomous triage and SAR drafting. Immutable audit for the examiner. MCP for the auditor. Open core, managed cloud or self-hosted, per analyst seat — no per-customer-screened tax.

Join the waitlistStar on GitHub