Documentation
Practical starting points for Dataverse practitioners evaluating DVQR: begin from Quick Starts or 94 guided prompts, talk to Dataverse through Local MCP, inspect the live Operational Profile + DVQR Score, build metadata-aware queries, traverse relationships, compare environments, reconstruct timeline evidence, generate Mini RCA briefings, and export investigation handoffs.
Quickstart
Step 1
Open the Prompt Library and choose a Quick Start, search one of 94 guided prompts, or work directly in the editor. Every route leads back to the same deterministic DV Quick Run capability surface.
Prompt Library: 94 prompts across six categories · 69 Free + 25 Pro. Local MCP: 30 Free tools + 32 Pro-only tools = 62 total with Pro.
Step 2
Install from the VS Code Marketplace, then open the DV Quick Run Hub from the command palette.
Command: DV Quick Run: Open Hub
Step 3
Use your existing Dataverse environment connection context. DVQR is designed to work where Dataverse practitioners already investigate, explain, and verify operational evidence.
Keep environment context visible before running operational workflows.
Step 4
Ask DVQR to explain or recommend a Custom API, then create a preview for an eligible public global Action. Review the plan, reply EXECUTE in a later message, and interpret the stored evidence afterward.
Recommended flow: discover → explain → recommend → architecture → preview → EXECUTE → interpret.
Step 5
Start with the smallest OData or FetchXML query that reproduces the operational question. For OData, even a bare entity set can begin metadata-aware lookup discovery.
Useful first surfaces: Show Metadata-Aware Query Suggestions, Explain Query, Result Viewer, Copy Query, Copy JSON.
Step 6
Use Result Viewer actions, Query Understanding, Cross Diff Explain, Access Context, Operational Profiles, or comparison workflows to continue investigation without treating signals as certainty.
DVQR observes operational drift. DVQR does not fix operational drift.
Investigation Playbooks
Use Custom API Intelligence when you need to understand an operation, preview an eligible execution, confirm it explicitly, and preserve the runtime result for interpretation.
Use Metadata-Aware Query Suggestions when you need a lookup identifier or target-specific navigation without memorising Dataverse relationship syntax.
Use Query Understanding when a query needs operational interpretation rather than clause-by-clause syntax notes.
Use Cross Diff Explain before detailed evidence review to understand what changed, why it matters, how confident DVQR is, and where to investigate next.
Use Timeline Reconstruction with 3+ compatible snapshots from the same entity and environment to understand when drift was first observed.
Use Mini RCA (Experimental) from Cross-Environment Diff or Timeline Reconstruction to turn canonical investigation evidence into a bounded report with Investigation Readiness, contributor candidates, deterministic Evidence Correlation, shared recommendations, and appendix-backed technical trace.
Use Cross-Environment Diff to compare operational participation, runtime behaviour, workflows, identity participation, relationship metadata, column metadata, choice metadata, entity configuration, environment variable current values, and reportable drift evidence.
Use eligible Metadata Attribute, Identity Participation, Choice Metadata, and Environment Variable Drift findings to export DVAF, DVIM, DVCE, or DVEVM reconstruction artifacts while keeping remediation and preview/apply outside DVQR.
Use plugin step, workflow, solution participation, and runtime drift surfaces to understand observed behavioural differences without claiming root cause.
Use Access Context to inspect bounded user, application user, team, role, and business-unit participation without simulating effective access.
Use report exports to produce portable evidence summaries that help humans verify next steps outside DVQR.
Operational Principle
DVQR surfaces operational evidence, comparison signals, and investigation continuity. It does not create deployment authority, remediation certainty, or autonomous root-cause claims. Humans remain responsible for validating operational decisions.
FAQ
Business Architecture Understanding is the GitHub Copilot experience powered by DV Quick Run's extension-owned Local MCP server. It can execute bounded OData GET requests, explain OData and lookups, inspect metadata, discover and rank relationship paths, generate verified relationship queries, probe bounded runtime evidence, and use the Custom API Intelligence lifecycle for discovery, explanation, recommendation, preview-confirmed execution, and interpretation.
Yes. The Local MCP foundation is included in Free, including read-only investigation, bounded OData GET execution, query explanation, metadata inspection, deterministic discovery, lookup explanation, navigation resolution, relationship-path discovery, query generation, basic Custom API Intelligence, and eligible preview-confirmed Custom API execution. Pro retains the commercial boundary for deeper intelligence, orchestration, environment awareness, evidence continuity, and productivity acceleration.
Open the DV Quick Run Hub and select Enable Local MCP. The preference is remembered per workspace. VS Code registers the server and starts the local stdio process on demand when its tools are discovered or used. You do not need to run a permanent service or keep a terminal open.
No reconfiguration is required. Enablement is remembered for the workspace across VS Code restarts and computer reboots. The stdio process itself is session-bound and VS Code starts it on demand in each session.
In v0.16.0, live Dataverse reads, managed investigation evidence acquisition, and eligible Custom API execution use an Azure CLI tenant session. Dataverse-only tenants can sign in with az login --allow-no-subscriptions. DVQR's normal editor workflows continue to use their existing connection context. Reusing the extension authentication session is planned as a later refinement.
Not generally. OData and investigation tools remain read-only. v0.16.0 retains the deliberately narrow POST path for eligible public global Custom API Actions only after a short-lived preview and a later explicit EXECUTE confirmation. PATCH, DELETE, upload, remediation, administrative operations, hidden mutation, and unpreviewed execution remain outside the Local MCP authority boundary.
No. The MCP server runs locally as a VS Code-managed stdio process. DVQR does not host the server or upload your Dataverse records, metadata search catalogue, snapshots, or investigation artifacts. As with any Copilot workflow, review your organisation's GitHub Copilot data policies separately.
v0.16.0 is packaged and verified for VS Code and GitHub Copilot through VS Code's MCP support. The server uses the standard MCP stdio protocol, but other clients are not yet presented as supported DVQR workflows.
No. DVQR is designed as a local-first operational investigation tool. Operational investigation data, Dataverse records, snapshots, exports, and comparison evidence are not uploaded during entitlement validation.
The Prompt Library is a guided discovery surface for DVQR's MCP capabilities. It contains 94 curated user-intent prompts—69 Free and 25 Pro—across six categories. Quick Starts help users begin from common outcomes; search, category and tier filters narrow the catalogue; required parameters render into the exact copy-ready natural-language request; and suggested next prompts continue the evidence-backed journey. The library is guidance and catalogue metadata, not a second reasoning engine.
Free MCP exposes the canonical live Operational Profile and DVQR Score through Free MCP. It uses the same profile/evidence/scoring path as the GUI, exposes weighted evidence and calibrated score math, preserves unavailable evidence as unknown rather than zero, and explicitly prevents DVQR Score from being interpreted as a health, risk, security, performance, quality, business-value, or root-cause score. For current-state questions, the live result is authoritative over older workspace snapshots unless the user explicitly asks for historical comparison.
Custom API Intelligence is DVQR's end-to-end metadata-to-runtime lifecycle: discover, explain, compare, recommend, architect, preview, explicitly confirm, execute, and interpret. Execution uses a short-lived single-use preview session. Interpretation reads stored runtime evidence and never reruns the operation or contacts Dataverse again.
No. Each preview is short-lived and atomically consumed before transport. A second attempt with the same preview is rejected, even when the first Dataverse call returned an HTTP error. A fresh execution always requires a fresh preview and a new explicit confirmation.
New in v0.16.0, Managed Business Paths let DVQR preserve an explicitly runtime-verified relationship route as workspace-owned Preferred traversal knowledge. Saved paths are inspectable JSON artifacts under .dvforgelab/dvqr/business-paths, surface first without hiding metadata-derived alternatives, and can be reused from a real source record through the exact saved relationship hops. Preferred means user-selected workspace knowledge, not universal business truth or causal proof.
New in v0.15.8, Managed Investigation Intelligence lets Pro users start from a real Dataverse record and continue a persisted MCP investigation across turns. DVQR preserves the subject, explicit downstream target, confirmed intent, deterministic strategy, acquired evidence, readiness state, Mini RCA checkpoint state, and state-consistency diagnostics. Deterministic actions are bound to exact tools and persisted-safe arguments through scoped action IDs; stale, replayed, substituted, or altered requests are rejected. Observed participation and runtime reachability remain evidence, not proof of mechanism, root cause, or causality.
Investigation Readiness explains whether the evidence supplied to Timeline or Cross-Diff Mini RCA is prepared for bounded synthesis. Ready, Conditional, Limited, and NotAssessable are qualitative postures—not scores. The assessment preserves contributor availability distinctions, evaluates provenance, coverage, freshness, scope, repeatability, and consistency, and may preserve or reduce synthesized confidence. It never certifies truth, proves root cause, raises confidence, or authorises remediation.
Mini RCA (Experimental) is an evidence-backed operational explanation layer. In v0.15.3, Cross-Environment Diff and Timeline Reconstruction flow through canonical investigation input and the same Investigation Readiness contract. Reports separate subject-local evidence from cross-cutting context, expose explicit dominance criteria, preserve provider-owned evidence, and explain material gaps without claiming root-cause certainty or causation.
DVQR uses metadata from the active Dataverse environment to recognise standard and polymorphic lookups, validate navigation properties, and preview target-specific OData suggestions. It separates schema-valid targets from the target used by a particular row, never guesses uncertain fields, and never executes a suggested query. Free users can inspect and copy; Pro users can explicitly apply a preview.
No. DVQR focuses on evidence-backed operational investigation. It can help structure review and handoff, but it does not claim autonomous diagnosis, root-cause certainty, or remediation authority.
DVQR is preview-first and evidence-first. Execution workflows are governed and explicit. Comparison and investigation surfaces are observational unless the user intentionally performs a supported execution action.
Free currently exposes 30 Local MCP tools for foundational Dataverse understanding, Managed Business Path discovery/validation, bounded queries, metadata and relationship intelligence, and eligible Custom API workflows. Pro adds 32 Pro-only MCP tools, for 62 total. The Pro MCP surface includes persisted Managed Investigation Intelligence: deterministic continuation, target-aware runtime evidence, readiness, bounded Mini RCA, state reconciliation, and investigation handoff, alongside the existing Pro comparison, timeline, audit, reporting, and reconstruction workflows. Online Pro plans include a 14-day free trial.
Pathfinder Founder is the limited founder pricing tier for the first 200 customers. It includes the same Pro capabilities, a 14-day free trial, and $19/month founder pricing for as long as the subscription remains active.
Offline Pro is annual-only, manually issued licensing for regulated, locked-down, and air-gapped environments. Commercial transactions still flow through DV ForgeLab and Lemon Squeezy.
Community
Feedback, operational scenarios, feature ideas, and roadmap discussion happen through GitHub. DVQR evolves through rapid iteration, dogfooding, and Dataverse practitioner feedback.
