Master Function & Feature Framework
This document maps the product/feature framework to METRC MCP capabilities. Use it to decide what to build in this repo (tools + skills) vs in your app (UI, rules engine, external data).
Legend: MCP = implement with current or new tools/resources. Skill = analysis pattern (LLM + tools). App = needs UI, rules engine, or external data.
| See also: Tools | Skills | Getting started |
1. BULK & EFFICIENCY TOOLS
1.1 True Bulk Inventory Actions
- Description: Bulk adjustments, status changes, package actions (not 1-by-1). Preview before execution.
- METRC feasibility: HIGH (array payloads).
- Implementation: MCP — Add tools like
metrc_adjust_package(already have), extend to accept arrays of adjustments; addmetrc_bulk_finish_packages,metrc_bulk_change_location. Optional tool:metrc_preview_bulk_adjustments(dry-run / validation). - Skill: “Bulk actions” skill: list packages to change → build payload → show preview (tool returns summary) → user confirms → execute.
2. INVENTORY AGING & LIFECYCLE AWARENESS
2.1 Intelligent FIFO / Aging Pull Recommendations
- Description: Recommend what to pull for samples, vendor days, sales kits, discounting. Use harvest/production/package age. Warn before breaking full case when partial exists, or pulling too fresh when older exists.
- METRC feasibility: HIGH (dates + quantities in packages/items).
-
Implementation: Skill + MCP. Skill: “What should I pull for [samples vendor days discounting]?” — call metrc_get_packages,metrc_get_harvests; sort/filter by date and quantity; apply FIFO logic in prompt or a small helper. App for brand-specific prioritization rules. - MCP: Implemented as skill FIFO / aging pull — see Skills. URI:
metrc://skills/fifo-aging-pull.
2.2 Aging Inventory Discount / Sampling Recommendations
- Description: Proactively identify inventory near brand/legal/shelf-life limits; best candidates for discount or sampling.
- METRC feasibility: HIGH for detection; shelf-life interpretation = App logic.
- Implementation: Skill for “What’s approaching age limits?” using package/harvest dates; App for configurable thresholds and rules engine.
3. INVENTORY INTELLIGENCE
3.1 Slow / Non-Moving Inventory Detection
- Description: Identify products/packages not moving; cash tie-up; operational risk.
- METRC feasibility: HIGH (sales + package endpoints).
- Implementation: Skill + MCP. Skill: “Show slow-moving or non-moving inventory” —
metrc_get_packages, optional sales data if available; LLM or post-process for velocity. App for trend/velocity store.
3.2 Future Compliance Risk Forecasting
- Description: Forecast expiration, shelf-life, structural (fragmentation, lab timing) risk.
- METRC feasibility: HIGH.
- Implementation: Skill for “Compliance risk forecast” using packages/harvests/lab; App for rules + forecasting engine.
4. EMPLOYEE & TRAINING RISK DETECTION
4.1 New Employee Pattern Detection
- Description: Detect new-hire mistake patterns, repeated adjustment errors, unusual vs team norms. Training needs; prevent compliance drift.
- METRC feasibility: MEDIUM–HIGH (sales have attribution; inventory actions limited unless proxied).
- Implementation: App — attribution and history; MCP can feed
metrc_get_*for context. Skill possible for “Review recent adjustments” if we have adjustment history.
4.2 Real-Time Mistake Warnings
- Description: Warn before wrong unit, wrong package, out-of-trend adjustments.
- METRC feasibility: MEDIUM (strongest when actions originate in your system).
- Implementation: App (overlay/co-pilot UI + risk scoring). MCP tools can be called to validate before commit.
5. INTEGRATION & DATA QUALITY
5.1 Cross-System Data Reconciliation
- Description: METRC ↔ Dutchie, Flowhub, LeafLink, Stashstock, ERPs — missing data, duplicates, out-of-trend, sync issues.
- METRC feasibility: HIGH on METRC side; full power with partner APIs.
- Implementation: App (multi-source reconciliation). MCP supplies METRC snapshot; Skill “Reconcile METRC vs [system]” if other system is queryable by LLM.
6. CUSTOMER & INVENTORY PRIORITIZATION
6.1 Early Warning for Key Customer Fulfillment Risk
- Description: Customer orders X every Y weeks; inventory may not cover next order; alert to prioritize production/prep.
- METRC feasibility: HIGH (historical sales + inventory).
- Implementation: Skill for “Will we have enough for key customers?” using packages + (if provided) order history. App for predictive logic and CRM data.
7. PACKAGE OPTIMIZATION & CLEANUP
7.1 Partial Package Fragmentation Detection
- Description: Detect when employees break new cases and ignore existing partials.
- METRC feasibility: HIGH.
- Implementation: Skill — “Show fragmentation” —
metrc_get_packages, group by item/location, flag multiple partials where one could be used.
7.2 Package Consolidation Recommendations
- Description: Recommend re-sticker, combine low-counts, simplify selling units.
- METRC feasibility: MEDIUM–HIGH (recommendation easy; execution state-dependent).
- Implementation: Skill for recommendations; App for execution workflow.
7.3 Sample-Out Low Counts for Sales Enablement
- Description: Identify low counts better used as samples/sales incentives.
- METRC feasibility: HIGH.
- Implementation: Skill — “Which low-count packages should we sample out?” using packages + quantity thresholds.
8. PRODUCT & BIOMASS UTILIZATION
8.1 Smart Biomass Utilization
- Description: Recommend best use of uneven biomass (don’t burn 6 lbs on fresh test; route small lots to extraction; large for pre-rolls).
- METRC feasibility: HIGH (quantities + lineage).
- Implementation: Skill using packages/harvests/items; App for strong proprietary logic.
8.2 Historical Yield-Aware Recommendations
- Description: Strain historically processed well but now used inefficiently; recommend better use cases.
- METRC feasibility: MEDIUM–HIGH (Metrc quantities; yield math = yours).
- Implementation: App (yield store); Skill can use strain + package data for context.
9. COMPLIANCE CONFIDENCE TOOLS
9.1 Audit-Ready Snapshots
- Description: “Audit in a week — check common risk areas.” Health snapshot, trend flags, areas to review.
- METRC feasibility: HIGH.
- Implementation: Skill — “Audit readiness check” — call harvests, packages, transfers, tags, inactive lists; summarize risks and cleanup items. App for trend + rules engine.
- MCP: Implemented as skill Audit-ready snapshot — see Skills. URI:
metrc://skills/audit-ready-snapshot.
9.2 Employee Self-Audit Mode
- Description: Employees check own work for errors, patterns, risk.
- METRC feasibility: MEDIUM–HIGH (attribution limitations).
- Implementation: Skill for “Review my recent actions” if we have action history; App for attribution.
10. TREND ANALYSIS (CROSS-CUTTING)
- Description: Out-of-trend behavior, historical anomalies, silent drift.
- METRC feasibility: HIGH.
- Implementation: App (core engine). MCP/skills feed data; Skill “Anything out of trend?” can run spot checks from current state.
11. TOLLING CALCULATOR
11.1 Toll / Split Agreement Financial Engine
- Description: Fees, splits, costs; yields biomass → finished; yield/loss by stage, inventory, financials.
- METRC feasibility: HIGH for quantities; costs in your system.
- Implementation: Skill for yield and inventory from METRC; App for costs and financial engine.
12. AI STRAIN & CROSS INTELLIGENCE
12.1 Strain Pairing & Product Fit Recommendations
- Description: Recommend crosses, blends, product types from inventory, history, goals.
- METRC feasibility: MEDIUM–HIGH (inventory context; creativity = yours).
- Implementation: Skill “Strain/product recommendations” using
metrc_get_strains,metrc_get_items,metrc_get_packages; App for AI/heuristics IP.
13. CUSTOM REPORTS
13.1 User-Defined Report Builder
- Description: Build, save, reuse reports from METRC endpoints and derived fields.
- METRC feasibility: HIGH.
- Implementation: Skill — “Build a report: [dimensions]” — LLM chooses tools and shapes output. App for GUI report builder and saved definitions.
14. YIELD ANALYZER
14.1 Stage-by-Stage Yield Analysis
- Description: Yield at raw material, production, test, final product.
- METRC feasibility: HIGH.
- Implementation: Skill for “Yield analysis” using harvests, packages, lab tests, items; App for yield modeling and storage.
Summary: MCP/Skill vs App
| Build in MCP/Skills (this repo) | Build in App (UI, rules, external data) |
|---|---|
| Bulk actions (array tools + preview skill) | Brand/shelf-life rules, trend engine |
| FIFO / aging pull recommendations (skill) | Real-time mistake overlay, attribution |
| Slow-moving detection (skill) | Cross-system reconciliation, velocity store |
| Fragmentation + consolidation + sample-out (skills) | Employee pattern detection, risk scoring |
| Audit-ready snapshot (skill) | Custom report builder GUI, saved reports |
| Yield / tolling from METRC data (skills) | Cost/financial engine, historical yield DB |
| Strain/product fit from inventory (skill) | Predictive customer fulfillment, AI strain IP |
Use this doc as a roadmap: implement skills and array-capable tools first for high-feasibility items; reserve App for proprietary logic, GUI, and multi-source data.
| ← Back to documentation | Tools | Skills |