Capital. Intelligence. Governance. One connected system.
Five parts of the QIS Ecosystem are live today, across three domains — not a roadmap, but a framework, a set of drafted instruments, an assessment, a validated schema, and the graph that ties them together. Live means a URL resolves. Inspectable means the method, the schema, or the source is published beside the claim, not asserted apart from it.
One system, three layers.
Every autonomous financial system has to answer three questions: who produces the intelligence, where the capital is deployed, and how accountability is enforced. The three domains below answer one question each, and the graph beneath them is what keeps the three answers from drifting apart.
QISAgent.com
Research · Construct · ExecuteThe intelligence layer — where markets are researched, strategies are constructed, and execution runs within defined parameters.
INSIDE · the QIS Atlas · the daily tape · the Agent Stack Enter the intelligence layer → Capital · The "Where"QISFund.com
Capital · Deployment · GovernanceThe capital layer — the recognized vehicle through which strategy becomes allocation, connecting intelligence to liquid markets.
INSIDE · strategy taxonomy · the glossary · the knowledge graph Browse the strategy taxonomy → Governance · The "How"QISTrust.com
Governance · Auditability · AccountabilityThe trust layer — oversight, auditability, and fiduciary accountability. The architecture that makes autonomous systems allocatable.
INSIDE · MARQUE · KYA · the QIS Canon · the Clause Library Enter the governance layer →What's actually running.
Each one is published, versioned, and open to use today — checked against the live build, not written up from a plan.
The 2026 interagency guidance on model risk (SR 26-2) places agentic systems outside its scope and directs institutions back to practices built for a model that holds still. MARQUE addresses delegated authority directly — how to give an autonomous system consequential authority and remain answerable for what it does. Free to adopt and citable by name. 8 questions · v1.0 · mapped against 19 published standards and 230 control objectives.
An authority grant, a kill-switch covenant, provenance representations, a stressed-conditions annex, and the schedule form that completes all four. Every provision cites the authority it derives from, so counsel can verify the derivation rather than take it on faith, and each instrument states the questions it does not resolve. 5 drafted instruments · derivations cited to SR 26-2, Rule 15c3-5, RTS 6, DORA, SMCR, ISDA and Basel.
Twelve dimensions, each scored on whether the evidence can be produced today, without preparation — not on whether a policy exists. The ladder is calibrated against production, not intention: the top of each dimension requires an artifact that already exists, which is what makes a strong score mean something to a diligence team. 12 dimensions · 5-level evidence ladder · free, no sign-up, nothing transmitted.
Every instrument in the library defers its hardest content to a schedule: the limits, the halt triggers, the thresholds. A published JSON Schema defines the governed-quantity record all of them rely on, and the Schedule Builder produces a valid instance in the browser, checked live against that exact schema. A public conformance suite means any independent implementation can be tested against it too — the record format is open, not proprietary. Live at qistrust.com/schedule-builder/ · schema and fixtures at qisfund.com/api/.
The capital, intelligence and governance layers reference each other directly through a machine-readable graph, a shared controlled vocabulary, and typed cross-links between every term, page and instrument. The structure resolves as data as well as prose — the graph is CORS-enabled and traversable by a system rather than only readable by a person, which is what lets the framework, the instruments and the assessment be used together rather than read separately. Since 2026-08-18 it is also published as formal RDF/OWL — the QIS ontology mapped to PROV-O and to one verified FIBO class where a genuine correspondence exists, not asserted wholesale — and checked against a public SHACL shapes file on every build, so conformance can be confirmed independently rather than taken on faith. Every node additionally exposes its own bounded neighborhood as a standalone document, sized for a retrieval system that needs one thing's context without the whole graph. Knowledge graph at api/graph.json · RDF/OWL at api/graph.ttl · ontology at api/ontology.ttl · SHACL shapes at api/shapes.ttl.
Published, versioned, built to be relied on.
The instruments are drafted to be executed against, the assessment is scored on evidence rather than intention, and the record format is specified precisely enough that conformance can be tested independently rather than asserted by us. Access to read and cite is not the same claim as a license to redistribute or train on the corpus — see the license for the distinction.
This page is an index, not a substitute for the pages it points to. The instruments, the assessment method, and the schema are each maintained on their own page, linked below — read those before relying on anything stated here.
Figures above are drawn from the live build and checked at time of writing, not asserted from a plan. The Clause Library is published as a form book with commentary — not legal advice, not a compliance determination, not a certification — and should be reviewed by counsel before use. Corrections and licensing inquiries: editor@qisfund.com.