GrantsProQR
Reference Implementation · Decision Lineage Protocol

The first DLP-conformant product.

GrantsProQR is the Decision Lineage Protocol running in a real domain — authority-anchored, go/no-go compliance intelligence for federal grants. Every determination carries citation lineage. Refusal is first-class. Professional judgment is never substituted.

Request early access →See how a decision moves through the system
Federal grants compliance2 CFR 200 & related requirementsGoverned under the GrytLabs world modelPre-product preview
What it is

Authority-anchored compliance intelligence — bounded, cited, and honest about its limits.

GrantsProQR bridges federal grants regulatory complexity to practitioner confidence by delivering authority-anchored, go/no-go compliance intelligence bounded at the federal level, so that grant-receiving entities can act with defensible speed on 2 CFR 200 and related requirements — without substituting professional judgment.

Federal grants fail not because the rules are unknown, but because authority is layered and conditional, applicability is contextual and time-bound, and the reasoning behind decisions is not preserved. Most systems flatten that complexity into documents and checklists. GrantsProQR makes it navigable — not by simplifying the complexity, but by composing it into decision pathways with full citation lineage.

Every query resolves to one of four classes

No undifferentiated “output.” Each determination carries an explicit type.

Confident go

The action is permitted under cited authority.

Confident no-go

The action is prohibited under cited authority.

Conditional interpretation

The action may be permitted contingent on named conditions, with explicit caveats.

Explicit refusal

The system cannot determine; it produces a gap statement with guided next steps.

What it is not

  • Not a chatbot, document summarizer, or search engine — a compliance-intelligence instrument with defined planes of reasoning.
  • Never a substitute for professional judgment — it supports human decisions; it does not render audit opinions or replace the grants professional, auditor, or legal counsel.
  • The federal confidence boundary is structural — state, local, and entity policy surface as “requires further consultation,” never as resolved conclusions.
  • A fast answer without traceable authority is worse than a refusal.
How it works

Watch one question move through the system.

A determination is not a lookup. It is a composition — context assembled, authority retrieved, prior decisions weighed, claims formed with explicit limits, gaps named and routed, and the whole reasoning trace captured so it can be replayed and defended. Press play, or step through each stage.

Illustrative walk-through. The entity, award, and requestor are fictional.

Decision walk-through · one question, seven stages
Context Assembly
What does the system know?Before reasoning begins, the system assembles everything relevant to this question.
Decision trigger
Can we charge the $12,400 professional development conference for our grants team to the CDBG-DR award?
City of RiversideHUD CDBG-DR 2024 (B-24-DF-12-0001)$12,400Conference registration closes in 5 days
Entity profile loaded
City of Riverside
Type: Municipal Government
Federal expenditure: $8.4M (FY2025)
Cognizant agency: HHS
2024 ProcurementResolved2023 Allowable CostsResolved
Award context retrieved
B-24-DF-12-0001
CDBG-DR (ALN 14.269)
Amount: $1,800,000
Period: 2024-01-15 to 2027-01-14
Davis-BaconEnvironmental ReviewDuplication of Benefits
Requestor profile
Maria Chen
Grants Coordinator · 2.3 years
Prior decisions
47
Decision quality
94%
Subject context
National Grants Management Association Annual Conference
Dates: February 15-18, 2026
Location: Washington, DC
Focus: Federal grants compliance, 2 CFR 200 updates, disaster recovery programs
CPE credits: 24
Context assembly complete
4 context objects loaded · 12 attributes indexed · ready for authority retrieval
Service modules & triggers

From signal to scoped response — composed, not guessed.

The same discipline that produces a determination composes a scoped response. A finding arrives as a typed signal — its domain, its questioned-cost exposure, the flags that make it what it is. Rules select the service modules that apply, at the depth the signal warrants. No model call. Every module carries the rule that placed it in scope — the lineage travels with the output.

Change the signal on the left; the scope recomposes deterministically. Illustrative modules and rules; internal effort and pricing bands are omitted.

Finding signal
Proposed scope · 3 modulesComprehensive control baseline
M1CAP Testability & EvidenceMicro
Turn corrective-action-plan promises into documented, testable controls with clear evidence paths.
CB outputs
  • Control statements
  • Evidence index
  • Test map
  • Control Data Map contribution
micro deliverables
  • 1–2 control statements
  • Evidence index
  • Test map (3–5 pages)
Reporting domain → CAP testability baseline
M2Governance & AccountabilityMicro
Establish clear ownership and oversight so fixes stick.
CB outputs
  • RACI matrix
  • Approval & escalation flow
  • Control calendar
micro deliverables
  • RACI
  • Control calendar
  • 1-slide governance snapshot
Repeat finding or cross-department scope → governance
M4Reporting & Financial ReconciliationMicro
Tie program data ↔ general ledger ↔ SEFA cleanly to stop reporting errors and questioned costs.
CB outputs
  • Lineage map
  • Quarter-pack checklist
  • Recon template + example
micro deliverables
  • Lineage map
  • Quarter-pack lite
Reporting domain → reporting & reconciliation
Glue artifacts
Control Data MapGap AnalysisAutomation Readiness Score
Assumptions
single-programno system reconfiguration
Guardrail
No assurance + implementation same period
Next step
Introduce a delivery agreement at baseline close only if proceeding to automation readiness.
The protocol underneath

GrantsProQR is one instantiation. The discipline is the Decision Lineage Protocol.

DLP is the open structural layer for AI-mediated decisions: typed truth, encoded authority, and preserved lineage. GrantsProQR is what it looks like when that layer runs against real regulation, for real practitioners, with real consequences.

Decision lineage

Every determination is captured as a replayable trace — context, authorities, claims, gaps, and the accountability chain. Not an audit trail after the fact; lineage carried at decision time.

Truth-type discipline

Every output carries an epistemic type: authoritative citation, derived determination, conditional interpretation, or explicit gap. The four truth types from DLP-Core, instantiated in the product's reasoning architecture.

Authority planes

Federal authority carries confident assertions; entity-specific context carries conditional interpretations. The confidence boundary is structural — not a setting, not a disclaimer.

Refusal-first

When the system cannot determine, it refuses and names the gap. The governance substrate is not optional infrastructure — it is constitutive of what the product is.

“A decision without lineage is a conclusion without evidence.”
— DLP quotable principle
What it offers

Governance-grade compliance intelligence, at a practitioner's price point.

It sits between boutique advisory and generic AI: domain-specific where general tools are shallow, citable and durable where chatbots are not, and priced for the entities that carry federal awards but cannot afford hyperscale consulting.

Natural-language query

Ask a compliance question in plain language, bounded to your entity type and program.

Four-class determinations

Confident go, confident no-go, conditional interpretation, or explicit refusal — never undifferentiated output.

Citation chains

Every confident assertion resolves to a specific, verifiable regulatory citation. Gaps are explicit refusals.

Technical-note export

Structured reasoning traces exportable for your compliance documentation and audit defense.

Entity + program scoping

Declare what kind of entity you are and which programs you run; the confidence perimeter is set for the session.

Guided gap resolution

When the answer depends on non-federal law, the system routes you to the next step — never a hollow conclusion.

Access tiers

A free tier anchors access; paid tiers add reach. Pricing published at launch.

Free

Entity + program selection, with at least one authority-anchored response per session.

Essentials

Single user, core determinations, technical-note export.

Professional

Team access, full feature set.

Connect

Bundled with GrytLabs advisory for entities that need the human loop.

Early access

Be first to run a determination.

GrantsProQR is in build. Join the preview list and we’ll reach out as access opens — for grant-receiving entities and the advisors who serve them.

In build. No spam — just an access note when the preview opens.