Overview: Gain Business Insights With the Canix MCP Server

Objective: Once you've connected the Canix MCP Server to Claude or ChatGPT (see our Setup & Getting Started Guide), you can ask natural-language questions about your Canix data and get instant answers — no exports or manual lookups required. This article shows the kinds of questions the server can answer today, grouped by data area, with example answers and visuals.

Benefit: This reduces time spent on reporting, surfaces insights you might not think to look for, and lets you combine data across your operation in a single query.

Business Logic: Each data area is queryable on its own or combined in a single question — so you can ask both targeted questions and cross-functional ones that no single report covers. Answer quality depends on the completeness of your underlying data.

The MCP Server currently exposes:

Cultivation: plants, plant batches, harvests, plant losses, cultivation costs, and harvest margin.
Inventory: current package-level stock, status, location, COGS, aging, and allocations. Plus inventory adjustments and losses, such as spoilage, waste, and samples.
Non-Cannabis Inventory (NCI): packaging, jars, labels, nutrients, and solvents: stock on hand, usage and costs, month-start and month-end balances, and transfers between facilities.
Lab Tests: compliance test results from Metrc, BioTrack, Confident Cannabis, and manual uploads: pass rates, retests, potency, terpenes, dosing, turnaround, and COA status.
Manufacturing: production batches and runs, inputs and outputs, yield, waste, recipes, work in progress, labor hours, and production cost.
Labor: hours logged against tasks and runs, labor cost, and labor cost per gram of output.
Purchasing: purchase orders, vendor details, costs, receiving status, and payment terms.
Sales: order line items, revenue, customers, sales reps, brands, discounts, and taxes.

More data types are planned for future releases.

Each section lists questions you can ask, with flagship examples showing what answers look like. You can ask any of these in plain language and follow up to drill deeper.

Note: The images in this article use sandbox or example data to show what answers look like. Every question was tested against the Canix MCP Server, and your results will reflect your own Canix data.

Glossary
Prerequisites & Setup
Explore By Data Area
System Impacts & Dependencies
How We Keep Answers Reliable
What the MCP Server Can't Do Yet
Troubleshooting Common Issues
Share Your Feedback

Glossary

COGS (Cost of Goods Sold): The total standard cost assigned to a package in Canix, comprising cannabis material, non-cannabis ingredients, and labor costs. Used to calculate inventory value and gross margin.

Domain: A data area the MCP server can query.

MCP (Model Context Protocol): The protocol that connects Canix to AI tools like Claude and ChatGPT, allowing you to query your Canix data in natural language.

NCI (Non-Cannabis Inventory): Supplies that aren't cannabis, such as packaging, jars, labels, nutrients, and solvents.

PAR level: The minimum quantity you want to keep on hand for a product before reordering.

Pass rate (first-pass and final): The first-pass rate counts tests that passed on the first try. The final pass rate includes retest results, so it can be higher or lower than the first-pass rate depending on how retests turned out.

Sales velocity: How quickly a product is selling over a given time period, typically measured in units or weight sold per day or week.

Yield: Total output divided by total input for a production run.

Prerequisites & Setup

Permissions: You must have a Canix account with access to the facilities and data you want to query. The MCP server enforces your existing Canix user permissions — you will only see data your account is authorized to view.

Necessary Data: Answer quality depends on how complete your Canix data is. For example:

For margin and cost analysis: COGS must be entered for your packages.
For open-order and pipeline views: Delivery dates must be set on your orders.
For purchasing analysis: Purchase orders must be recorded in Canix with vendor and cost details.

Related Resources:

Canix MCP Server - Setup & Getting Started Guide

Explore by Data Area:

Cultivation

The cultivation area covers plants, plant batches, harvests, losses, and cultivation costs.

Plants & Harvests

Try: "Which strains are producing the highest yield?"

The assistant ranks strains by average harvest weight, so you can see which genetics deliver the most.

More plant and harvest questions to try:

How many plants do we have in each growth phase right now?
What share of our plantings become tagged plants?
What is our total wet weight through to packaged weight?

Losses & Cost

Try: "Where in the grow cycle are we seeing the most plant losses?"

The assistant counts losses at every stage, including young plants lost before they were tagged, so the earliest (and often largest) losses aren't missed.

Inventory

The inventory area covers current package-level stock: status, location, COGS, aging, and allocations.

Stock Levels

Try: "Which products have less than 10 units in stock?"

The assistant lists low-stock products with their available quantities, so you can catch shortages before they become stockouts.

Aging & Compliance

Try: "Which packages are closest to expiration?"

The assistant surfaces packages by expiration or age, which helps you prioritize what to move first.

More aging and compliance questions to try:

Show me packages that have been in stock the longest.
Which packages haven't been submitted to compliance yet?

Cost & Value

Try: "What's the total standard cost of our current inventory?"

The assistant totals cost of goods across on-hand packages, and you can break it down by brand or item.

Non-Cannabis Inventory

The non-cannabis inventory area covers supplies like packaging, labels, nutrients, and solvents: stock, usage, costs, and transfers.

Stock & Planning

Try: "Which non-cannabis products are below their PAR level?"

The assistant lists supplies that have dropped below your restock level, so you can reorder before production is held up.

More stock and planning questions to try:

What's my total non-cannabis inventory by facility?
How fast am I burning through each non-cannabis product, and how many weeks until I run out?
Which products will run out before their vendor's lead time if I order today?

Costs & Usage

Try: "Which non-cannabis materials cost us the most in production?"

The assistant ranks your supplies by the cost they added to production, so you can see which materials drive your spend.

Lab Tests

The lab tests area covers every compliance test result your company holds, whether it came from Metrc, BioTrack, Confident Cannabis, or a manual upload.

Pass Rates

Try: "What is my compliance testing pass rate?"

The assistant reports both your first-pass rate and your final pass rate (including retests), so you can see how retesting affected your results.

Potency & Coverage

Try: "What are my highest potency strains?"

The assistant ranks strains by average potency across their test results.

More potency and coverage questions to try:

What is the typical delta-9 THC dose per serving in our edibles?
How long does it take from the test date to the result being released?
What share of our test results have a COA on file?

Note: Potency varies a lot by product type. For fair comparisons, ask within one category, such as "flower" or "vape cartridges."

Manufacturing

The manufacturing area covers production batches and runs: what went in, what came out, yield, waste, labor, and cost.

Yield & Throughput

Try: "What is our yield by processing stage?"

The assistant returns yield the same way the app calculates it, so you can compare stages and spot underperforming runs.

More yield and throughput questions to try:

What was our extraction yield last quarter, actual versus planned?
How long does each stage of processing take, and where are our bottlenecks?
How much material is tied up in unfinished runs right now?

Traceability & Cost

Try: "Ingredient lot [lot number] failed our QA review. Which produced goods contain it?"

The assistant traces a lot to the packages and intermediary products made with it.

More traceability and cost questions to try:

How much production waste did we have last quarter, by stage, in weight and cost?
Did last month's [recipe name] runs use the ingredient quantities the recipe planned?
What was our total production cost by facility last quarter, excluding labor?

Note: Lots that Canix picked automatically can be traced for runs submitted from about February 2026 onward. Lots chosen by hand can be traced further back.

Labor

The labor area (Task Management in Canix) covers hours logged against tasks and runs, and the labor cost Canix records.

Try: "Which facilities spend the most on labor?"

The assistant ranks your facilities by labor cost, with the hours behind it, so you can see where your labor dollars go.

Sales

Revenue

Try: "Show me monthly revenue for the past 12 months."

The assistant returns a month-by-month revenue total and can chart the trend, so you can spot seasonality and growth at a glance.

Example: monthly revenue over 12 months (sandbox data)

Customers

Try: "Who are our top 10 customers by revenue this quarter?"

The assistant ranks customers by revenue and can add order counts and average order value alongside, so you see both your biggest spenders and how often they buy.

Products, Brands, and Strains

Try: "What are our top-selling brands by revenue this quarter?"

The assistant breaks revenue down by brand, and you can pivot the same question to product, category, or strain.

Sales Representatives and States

Try: "Compare sales reps by revenue and order count this year."

The assistant returns a per-rep breakdown of revenue, orders, and average order value, which helps you see who drives high-value deals versus high-volume relationships.

More sales rep and state questions to try:

Show me shipped revenue by sales rep and brand for the last quarter.
Which states generate the most revenue?
Break down revenue by customer state for this year.

Open Order Pipeline

Try: "What's the total value of my open sales orders?"

The assistant sums committed-but-unfulfilled orders by status, so you can see what's in the pipeline. You can also ask it to pull data from orders with only one specific status.

Purchasing

The purchasing area covers purchase orders: vendor details, costs, receiving status, and payment terms.

Try: "Who are our top vendors by spend this year?"

The assistant ranks vendors by total purchase order spend and can break down by product category and order status, so you can see both your largest suppliers and their order patterns.

Combining Data Areas

These questions draw on more than one data area at once.

Fast Movers, Stock, and Suppliers

Try: "Which products are selling fastest — are we stocked, and who do we buy from?"

The assistant combines sales velocity (how fast a product is selling) with current available stock, then flags which fast movers are well stocked, running tight, or already out. Purchasing history surfaces who supplies it and when you last ordered. This is a reorder-planning view built from three data areas in one question.

Customer Profitability

Try: "Which customers generate the most margin, not just revenue?"

The assistant pairs sales revenue with inventory COGS to estimate gross margin per customer, revealing that your top-revenue accounts aren't always your most profitable.

Yield and Potency (Cultivation + Lab Tests)

Try: "Which of our highest-yielding strains make our most potent vape cartridges?"

The assistant pairs each strain's harvest yield with its lab results for one product type, so you can see which strains deliver on both.

Ingredient Recall (Non-Cannabis Inventory + Manufacturing)

Try: "Ingredient lot [lot number] failed our QA review. Which produced goods contain it?"

The assistant follows a supply lot through production to every package made with it.

System Impacts & Dependencies

Action	Dependency	Description
Querying sales revenue	Order status	Results vary based on whether you filter by shipped, open, or all orders. Specify status for accuracy.
Querying inventory cost	COGS data entry	Cost totals only reflect packages with a standard cost entered in Canix.
Querying NCI reorder needs	PAR levels & lead times	Below-PAR and runout answers rely on PAR levels and vendor lead times being set.
Querying planned vs. actual yield	Output targets	If targets were never set on runs, planned and actual will look identical.
Querying labor cost	Labor costing & hourly rates	Facilities not using labor costing show no cost. Workers without an hourly rate are booked at $0.
Querying open order pipeline	Delivery dates	Pipeline views rely on scheduled delivery dates being set on orders.
Querying purchasing data	PO records	Vendor and spend analysis requires purchase orders to be recorded in Canix with cost details.
Cross-domain queries	Data completeness	Results are only as accurate as the underlying data across each area involved.

How We Keep Answers Reliable

AI assistants don't always answer the same question the same way. That's why we test how the AI uses the Canix MCP Server, not just whether the server itself works.

We do this with automated evaluation tests. Each one asks the AI a realistic customer question, like the ones in this article, and then checks the answer:

Did it look in the right place for your data?
Did the numbers come from your actual Canix records, not guesses?
Did it follow Canix's rules, like leaving samples out of on-hand inventory and keeping grams and units separate?
Did it tell you when it couldn't answer, instead of making something up?
Was the answer actually useful?

When an evaluation test fails, we improve the guidance the server gives the AI and test again until we get consistent results.

New data areas are tested this way as they're built, and we re-run the tests when we make changes, so new features don't quietly change answers you already rely on.

Evaluation tests make answers more reliable, but no AI tool is perfect. If an answer looks off, ask the assistant how it got the number, or contact our support team.

What the MCP Server Can't Do Yet

To set expectations, some questions fall outside the current data scope:

Grow planning and yield forecasts: Plant counts and active harvests are now available, but forward-looking grow plans and yield forecasts are not yet queryable.
Production schedules: Canix doesn't have a production schedule to report on yet. You can ask about past runs and runs that are currently open.
Sales commissions: the server does not calculate commissions or track payouts. You can, however, pull the revenue inputs a commission plan uses (for example, shipped revenue by rep and brand) and apply your own rates outside Canix.
Sales forecasting: the server reports committed and historical data, not probability-weighted projections.
Which lab ran a test: This isn't recorded, so lab-specific questions can't be answered.
Individual terpene values: Only total terpenes are available, with the top three terpenes shown as a summary.
Shifts: Canix doesn't track shifts, so labor-by-shift questions can't be answered.
Bucking labor: This isn't currently recorded in reporting.
NCI beginning and ending balances before September 4, 2026: These aren't available. Usage and transfer history for earlier periods is still available.

💡 Troubleshooting Common Issues

The Setup & Getting Started Guide covers the basics (be specific about time ranges, name your facility, ask follow-up questions). A few more tips that help with the questions above:

Query returns unexpected results: Check whether you specified an order status — "shipped revenue" counts only fulfilled orders, while "open orders" counts committed-but-unfulfilled ones. Try adding "shipped" or "fulfilled" to narrow results.

Average order value looks higher than expected: Average order value is total revenue divided by the number of distinct orders, so a few large bulk orders can raise the average significantly.

Cost data looks incomplete: Inventory cost totals only include packages that have a standard cost entered in Canix. Keeping cost data current makes these answers more accurate.

Open order pipeline doesn't match expectations: Confirm that delivery dates are set on your orders. Pipeline views by date rely on scheduled delivery dates being recorded.

Purchasing data is missing or incomplete: Check that your purchase orders are recorded in Canix with vendor and cost details entered. The server can query historical purchasing data.

A generated visual has formatting issues: Minor issues like overlapping text or uneven spacing can occasionally appear. You can ask the assistant to correct these directly — for example, "can you give the text more space in the status badges?" or "the column headers are overlapping, can you fix that?" — and it will adjust and regenerate the visual.

A question doesn't return what you expected: Try rephrasing with more specifics — add a time range, facility name, or order status. If the issue persists, contact our support team or submit feedback here.

Cost figures missing from an answer: You may not have the permission needed for that cost (e.g. labor cost or NCI cost). Ask your Canix admin.

My totals look different than expected: By default, the assistant leaves demo and sandbox facilities out of company-wide totals. Ask which facilities were included, or ask about a specific facility by name.

Pass rates look different than expected: The first-pass rate counts only first tests. The final pass rate includes retests, so it can be lower if retests failed. Results without a pass/fail status, such as manual uploads, are left out of both.

We're just getting started

The MCP Server is new and evolving, and this list of questions will grow as we add more data areas. If a question doesn't return what you expected, or you'd like to see a capability that isn't here yet, we'd love to hear from you. Your feedback directly shapes what we build next.

Have a question or want to share feedback? Contact our support team or submit feedback here.

Questions the Canix MCP Server Can Answer

Overview: Gain Business Insights With the Canix MCP Server

Table of Contents

Glossary

Prerequisites & Setup

Explore by Data Area: