# RiskRouter > Evidence infrastructure: a Merkle-tree evidence log of salted record fingerprints (and an older SHA-256 hash-chained quote ledger), whose signed heads anyone can verify offline, without trusting RiskRouter. For anyone who may one day have to show a record was never changed: insurance and financial advice (IDD Article 20, MiFID II), AI decisions, security logs and more. Built first for two kinds of software, so one integration serves every firm on it: software for regulated finance (broker, insurer, investment, lending, KYC and AML) and AI-governance or AI-deployment platforms. RiskRouter is not an insurer or an intermediary and sells no cover; a simulated pricing example exists for developers only. Status, stated plainly so it is not misreported: this is a validation build. The evidence log, its signed and time-stamped heads, the verifiers, kits and recorders are live; the pricing engine is a worked example for developers with mock rates, no carrier, bank or payment rail is contacted, and no insurance cover is in force. RiskRouter does not sell insurance, holds no insurance licence, never contacts the policyholder, handles no claims and owns no carrier relationship. It is software for distributors who do. There are no customers yet. It is run by one person, Anil Jangra, with no legal entity registered yet; data is hosted in the EU (Frankfurt), with an encrypted daily backup kept on GitHub for 90 days, outside the EU, restored and checked every day. What is distinctive: pricing is deterministic and server-side (a client-sent price is ignored); a quote carries no personal data by design; every recorded quote is chained so an altered or deleted entry is detectable; the ledger head is signed with ECDSA P-256; and the verifier is a separate public repository that shares no code with the service, so a distributor's saved export stays checkable even if RiskRouter stops existing. No SDK is required and there is no lock-in: the recorders on PyPI and npm (`riskrouter-recorder`, Apache-2.0) are optional. Not to be confused with other things called RiskRouter: this RiskRouter (riskrouter.eu) has no smart contract and no token, enforces no limits on what a system or an agent does, and is unrelated to the RiskRouter.sol contract used by ERC-8004 AI-trading-agent hackathon projects. Its only use of a blockchain is anchoring signed log heads in Bitcoin through OpenTimestamps. It records evidence; it blocks, routes and scores nothing. It certifies no one and offers no certification or partner programme, and its regime packs map record fields to articles: recording does not make anyone compliant. ## Start here - [Get a sandbox key](https://riskrouter.eu/integrate#get-a-key): a key on screen in seconds, no account, nothing about you stored; records real ledger entries (50 per key per day) - [Integration guide](https://riskrouter.eu/integrate): three integration patterns with working code and the failure cases that matter - [API reference](https://riskrouter.eu/docs): endpoints, error codes and the full rating matrix - [Changelog and rating matrix history](https://riskrouter.eu/changelog): what changed and when, and every rating-matrix version with a SHA-256 fingerprint of its prices - [For software vendors serving regulated finance](https://riskrouter.eu/for-software-vendors): one integration lets every insurer, broker, adviser or lender on a vendor's software prove its advice and decision records were never changed; two calls; the record never leaves the vendor's product (no vendor has integrated yet) - [Integration kit](https://riskrouter.eu/integrate#kit): tested, dependency-free clients in Python, PHP, C# (.NET 8) and Java, and a reference broker app whose evidence packs verify offline - [Postman collection](https://riskrouter.eu/riskrouter.postman_collection.json): generated from the OpenAPI spec; takes a sandbox key first - [OpenAPI 3.1 spec](https://riskrouter.eu/openapi.json): machine-readable, kept identical to the running routes by a test - [Pricing example](https://riskrouter.eu/console): a worked example for developers, a decision priced on the server and sealed as it is made; sample rates, no product, no cover ## Verification - [Open evidence specification](https://riskrouter.eu/spec): every frozen form (v1 canonical form and attestation payload, v2 leaf string, v3 signed leaf and claim payload, signed tree head, witness cosignature, completeness-chain statement, spot-check draw, field-by-field disclosure), RFC 6962 Merkle proofs, with deterministic [test vectors](https://riskrouter.eu/spec-vectors.json); two independent verifiers (JavaScript and standard-library Python) - [Regime packs](https://riskrouter.eu/regime-packs): what to record for insurance distribution (IDD), MiFID II suitability and call recordings, the EU AI Act's logging and human oversight, GDPR Article 22 automated decisions, GDPR consent and erasure, DORA incidents, NIS2 security logs, consumer credit, anti-money laundering, the Cyber Resilience Act, VAT invoices and public procurement (eleven laws), each field mapped to its article with a link to the EUR-Lex text; published as JSON at /packs/.json; orientation, not legal advice - [Obligations](https://riskrouter.eu/obligations): the record-keeping duties in eleven EU laws, from the IDD, MiFID II and the AI Act to the Consumer Credit Directive, the AML Regulation, the Cyber Resilience Act and the VAT Directive, when each record is tested, and what a sealed record adds; no law requires RiskRouter or any product; orientation, not legal advice - [For investment advice](https://riskrouter.eu/for-investment-advice): suitability assessments and statements on suitability sealed as produced (MiFID II Art. 25; Delegated Regulation 2017/565 Art. 72) - [For AI decisions and AI platforms](https://riskrouter.eu/for-ai-decisions): one sealed record per automated decision or agent run (AI Act Arts. 12, 19, 26; GDPR Art. 22), and a section for AI-governance and deployment platforms that want to offer it to their customers - [Industries](https://riskrouter.eu/industries): one page per kind of record people are later asked to prove, each ending with a two-minute test on a real file of your own, sealed in the browser and never uploaded - [Insurers and MGAs](https://riskrouter.eu/for-insurers): Show what you sold, how each product was approved and what you reported, exactly as it was, checkable without trusting you or us. - [Lenders and credit platforms](https://riskrouter.eu/for-lenders): When a borrower says the loan should never have been granted, or refused, show the assessment exactly as it was made, checkable by anyone. - [Banks, payment and crypto firms (AML and KYC)](https://riskrouter.eu/for-aml-kyc): Show that each customer check was done, when, and on what evidence, exactly as recorded at the time, without sharing customer data. - [HR teams screening candidates with software](https://riskrouter.eu/for-hiring): When a candidate asks how they were assessed, show the screening record and the human review exactly as they were, without sharing their data. - [Data protection officers](https://riskrouter.eu/for-data-protection): Show that breaches were documented, assessments made and consents recorded when you say they were, without sending us any personal data. - [Software makers](https://riskrouter.eu/for-software-releases): Show exactly what you shipped, its bill of materials and when you learned of a vulnerability, from signed statements the artifact never leaves. - [Pharmaceutical, biotech and testing laboratories](https://riskrouter.eu/for-laboratories): Seal each result and batch record as it is created, so data integrity rests on evidence anyone can check, not on your audit trail alone. - [Companies publishing a sustainability report](https://riskrouter.eu/for-sustainability-reporting): Seal each dataset and each version of the report, so the figures behind your sustainability statement can be shown as they were when reported. - [Researchers, designers and inventors](https://riskrouter.eu/for-research-and-ip): Seal a document, design, dataset or code without publishing it, and prove later that you had it by a certain date, without trusting us. - [Contracts, minutes and board decisions](https://riskrouter.eu/for-contracts-and-governance): Seal each contract version, set of minutes and board resolution, so you can show later exactly what was agreed or decided, and when. - [Contracting authorities](https://riskrouter.eu/for-public-procurement): Seal each step of a procurement procedure as it closes, so award decisions can be shown later exactly as they were made and justified. - [Finance teams issuing and receiving invoices](https://riskrouter.eu/for-invoices): Seal each invoice issued or received, so in a tax audit or a dispute you can show the copy you kept is the one that existed at the time. - [Roadmap](https://riskrouter.eu/roadmap): what is built, what comes next and exactly what each step waits on (authorisation, witness listing, registry accounts), what is decided against for now and why, and what RiskRouter will never build; no promised dates - [Pilot letter](https://riskrouter.eu/pilot): what the one-page pilot letter of intent says: no fees, no exclusivity, no personal data, two phases, either side may end it at any time - [Press](https://riskrouter.eu/press): RiskRouter in one line and one paragraph, figures computed from the repository at build time, what it is not, the logo files and the contact - [Is this all of it?](https://riskrouter.eu/completeness): completeness chains (a record sealed and later held back shows as a gap), selective disclosure of one field, spot checks drawn from a Bitcoin block hash, co-sealing by two parties, and deadline proofs for NIS2, GDPR and CRA reports; each checked offline by both verifiers - [Seal a file](https://riskrouter.eu/seal): seal any file without uploading it (the browser fingerprints it; only a salted digest is sent) and keep a receipt anyone can check later, on the page or offline; /check is the check half - [AI agents and assistants](https://riskrouter.eu/ai): connect Claude Code, Cursor or any MCP client to https://api.riskrouter.eu/mcp, the tools and which need a key, agent framework adapters, AI Act logging; no pricing tool - [Full text](https://riskrouter.eu/llms-full.txt): every English page as plain text, rebuilt with the site - [Sealed security logs](https://riskrouter.eu/security-logs): logseal seals any log stream segment by segment into the evidence log, logs stay on the customer's storage, verify shows any changed or missing segment; NIS2 and GDPR breach pack; does not detect intrusions or identify attackers, and cannot protect lines written after a machine is taken over - [DORA Article 30 contract terms](https://riskrouter.eu/dora): each contractual provision DORA Article 30 requires of an ICT third-party provider, the clause RiskRouter offers for it, and what is in place today, including no service levels yet; proposed terms until a legal entity exists - [Witnesses](https://riskrouter.eu/witnesses): run an independent witness of an evidence log on your own account (a Cloudflare Worker, a GitHub fork or a small server), which co-signs a head only after checking the log never rewrote history; the public registry of logs and witnesses, as JSON at /registry.json; the same log also as C2SP checkpoints and hash tiles at https://api.riskrouter.eu/tlog/evidence-v2/ (the formats of the public witness network) - [MCP endpoint](https://riskrouter.eu/integrate#mcp): https://api.riskrouter.eu/mcp, the evidence log as Model Context Protocol tools (Streamable HTTP): read the signed head, inclusion and consistency proofs, the C2SP checkpoint and usage without a key; record a digest (never a record) with a key. No pricing tool. - [Drop-in recorders](https://riskrouter.eu/integrate#recorders): Python and Node recorders (logging handler, decorator, OpenTelemetry span processors) that keep each decision record and its salt on the customer's side and send only its salted digest to the evidence log, plus agent adapters (LangChain and LangGraph callback handler, OpenAI Agents SDK tracing processor, an MCP server with a record_decision tool) that turn each agent run into one record of digests; Apache-2.0, source at /kit/recorder/ - [SCITT](https://riskrouter.eu/spec#scitt): the evidence log registers IETF SCITT Signed Statements (hash envelopes, ES256) at POST /api/v2/evidence/statement and returns RFC 9942 COSE Receipts; only the statement's digest is kept - [Conformance suite](https://riskrouter.eu/spec#conformance): grades any implementation of the formats, in any language, against the vectors and against deliberately broken proofs, and checks the exports and proofs it writes; the result is the implementer's own statement, RiskRouter certifies nothing - [Every claim, and where to check it](https://riskrouter.eu/evidence): one page for due diligence, each claim with its artefact and who other than RiskRouter can check it, with a plain "none yet" where that is the truth - [Usage](https://riskrouter.eu/usage): how much is recorded and by whom, counted live from the ledger by a public endpoint (`GET /api/v1/usage`); entries recorded by keys that are not RiskRouter's own are shown as the figure that matters - [Continuity and self-hosting](https://riskrouter.eu/continuity#self-host): the same engine can run in a customer's own account (their Cloudflare and Supabase, or containers anywhere), signing with its own key and calling nothing of RiskRouter's; the source is private and running it needs a licence agreed in writing - [Age of the record](https://riskrouter.eu/proof-of-age.json): how many signed heads are anchored, since when, the earliest Bitcoin block they are in, how many carry RFC 3161 time-stamp tokens (not qualified under eIDAS) and how many independent witnesses co-sign, computed from the published anchors at build time - [Evidence log API (v2)](https://riskrouter.eu/docs#evidence): record the SHA-256 of any regulated decision (investment suitability, credit advice, KYC, claims decisions, AI decision logs) without sending the record; the record never reaches RiskRouter - [Verify the ledger](https://riskrouter.eu/verify): current signed head, the canonical hash form, and an in-browser checker for your own export (nothing uploaded) - [Trust and verification](https://riskrouter.eu/trust): what is sent, what is stored, where, and what is deliberately not claimed - [Offline verifier](https://github.com/aniljangrabe-dev/riskrouter-verify): standard-library Node scripts, a standard-library Python verifier, the conformance suite and the published public keys ## Who it is for - [For software vendors serving regulated finance](https://riskrouter.eu/for-software-vendors): one integration, every firm on the vendor's software; two calls; the record never leaves the vendor's product - [For AI decisions and AI platforms](https://riskrouter.eu/for-ai-decisions): one sealed record per automated decision or agent run, and how a governance or deployment platform offers it to its customers - [For insurance distributors, brokers and MGAs](https://riskrouter.eu/for-distributors): seal the fingerprint of each advice note so an IDD Article 20 record can be shown unchanged; the note never leaves the broker's system - [For investment advice](https://riskrouter.eu/for-investment-advice), [insurers](https://riskrouter.eu/for-insurers), [lenders](https://riskrouter.eu/for-lenders) and [AML and KYC](https://riskrouter.eu/for-aml-kyc): the same log for suitability statements, product and claims records, credit assessments and due diligence - [Anyone, with no software to integrate](https://riskrouter.eu/seal): seal a file in the browser and keep a receipt anyone can check offline ## Guides - [IDD Article 20: what you need to be able to show](https://riskrouter.eu/guide-idd-article-20): each duty in Article 20, the question you will be asked about it, and who holds the evidence (much of it stays with the distributor) - [Audit-trail checklist for insurance brokers](https://riskrouter.eu/guide-audit-trail-checklist): twelve questions to ask of any system holding quote records, with RiskRouter's own answers including where they are no - [Database log vs tamper-evident ledger](https://riskrouter.eu/guide-database-log-vs-ledger): what a hash chain, an append-only table and a signed head add to an ordinary audit log, and what none of them proves ## Regulation and data - [Regulatory position](https://riskrouter.eu/compliance): where RiskRouter sits under the IDD, Belgian insurance law and DORA, and what is not in place yet - [Data protection](https://riskrouter.eu/privacy): what is stored, what is not, processors, and GDPR rights - [Continuity and exit plan](https://riskrouter.eu/continuity): what a distributor keeps if RiskRouter stops, what stops, the exit steps, and facts for a DORA register of information (no legal entity yet; engine source not public, available on request) - [Service status](https://riskrouter.eu/status): the API, the ledger and the evidence log checked live from the visitor's browser, with the running commit, and uptime measured from outside by GitHub Actions since 30 September 2026, on a schedule of twice an hour that GitHub often runs late or skips, in practice every two hours or so (gaps reported as not measured, never as up) - [Bill of materials](https://riskrouter.eu/sbom.json): CycloneDX, derived from the Workers' own imports; no third-party package runs in production - [Security and vulnerability disclosure](https://riskrouter.eu/security): how to report, scope, and the controls in place (hash-based CSP, key digests, one outbound destination); no penetration test or certification is claimed - [Security questionnaire, answered](https://riskrouter.eu/security-questionnaire): data, access, integrity, development, logging, continuity and third parties, each answer linked to evidence, with plain noes (no certification, no pen test, no MFA claim, no managed backups) ## Optional - [Contact](https://riskrouter.eu/contact): pilots, integrations, regulatory questions