MCP FAQ (Claude and ChatGPT)

Updated 2 hours ago by Jason

What is the AROYA MCP?

The AROYA MCP lets you ask questions about your facility in plain English inside an AI assistant such as Claude or ChatGPT, and get answers drawn live from your AROYA data. MCP stands for Model Context Protocol, an open standard that lets AI assistants connect securely to outside data sources. AROYA hosts the connection for you; you add it to your AI assistant once and sign in with your AROYA account.

Who is it for? Anyone with an AROYA login: growers, cultivation directors, operations and executive teams. If you can see it in AROYA, you can ask about it.

What does it cost? It is free during the introductory period. Pricing is expected to be introduced with renewals in calendar year 2027; your account manager will share details before anything changes.

Is it read-only? Yes. Release 1 reads your data and never changes it. It cannot edit schedules, create tasks, or adjust setpoints. Write capabilities are being evaluated for a future release.

Which AI assistants are supported? Claude and ChatGPT are officially supported and tested. The same connection may work in other MCP-compatible tools, but AROYA Support can only help with Claude and ChatGPT.

Getting connected

What do I need before I start? Three things: an active AROYA login, an AI assistant plan that allows connectors (see the table below), and the MCP feature enabled for your facility. If you are not sure the feature is on, ask your AROYA account manager or Support to enable it.

How do I add the connector? In your AI assistant's connector or integration settings, add a new custom connector, give it a name (for example "AROYA"), and paste the server address from your setup instructions. When prompted to sign in, use the same email address you use to log in to AROYA. Step-by-step instructions available via this link.

Which plans support connectors?

How do I know it worked? Open your assistant's list of connected apps and confirm AROYA appears. Then start a chat and ask: "Which AROYA facilities can I see?" If it lists your facilities, you are connected.

Can my whole team use it? Yes. Each person adds the connector to their own AI assistant and signs in with their own AROYA email. Everyone sees only what their AROYA permissions allow.

What can I ask?

What data can it see? Nearly everything in your AROYA account that you are permitted to view: substrate and climate readings (water content, pore EC, temperature, humidity, VPD, light), irrigation events and schedules, rooms, zones and harvest groups, target ranges, alerts, journal entries, tasks, harvest records, and device and gateway status. The one notable exception is images: photos attached to scouting notes, tasks, or SOPs are not available.

What kinds of questions work well? Specific questions about a room, zone, or harvest group over a defined time window. Some examples to start with:

• "What is the current water content and EC in Flower 4?"
• "How many irrigation shots did Zone 2 in Veg 1 receive yesterday, and what was the total volume?"
• "Show me the overnight dryback for each zone in Flower 2 over the last 7 days."
• "Which rooms had VPD outside their target range this week?"
• "Compare average pore EC between Flower 1 and Flower 3 for the current harvest group."
• "List any sensors or gateways that have gone offline in the last 24 hours."
• "Summarize this week's journal entries for Flower 4."

Can it help with bigger analysis? Yes. Once the basics work, try asking it to compare harvest groups, find which irrigation or climate variables moved alongside yield, or write a daily morning summary for one room. For heavier analysis, use your assistant's most capable model (for example Claude Opus) and keep the question scoped to a few rooms and a defined period.

Can I automate reports? If your AI assistant supports scheduled tasks (for example Claude Cowork), you can schedule a prompt such as "Every morning at 6 AM, summarize overnight conditions and irrigation for Flower 1 through Flower 4" and have it delivered to you automatically.

Can I combine AROYA data with other systems? Yes. If your environmental controls, ERP, or compliance system also offers an MCP connector, you can add both to the same assistant and ask questions that span them, for example relating fan speeds or room setpoints to substrate readings.

Permissions, access and security

Who can see what? Exactly what they can see in AROYA today. The MCP enforces your existing AROYA user permissions, including facility membership and harvest group access. A user whose AROYA role only covers tasking will only be able to ask about tasking data.

How does sign-in work? Through secure OAuth sign-in using your AROYA email. Your AROYA password is never shared with the AI assistant, and the connection can be revoked at any time from your assistant's connector settings or by your AROYA administrator.

Does the AI company get a copy of our data? The assistant retrieves only the data needed to answer each question, in the moment you ask. AROYA does not bulk-export your data to Anthropic or OpenAI. How your assistant stores conversation history depends on your plan and settings with that provider; Team and Enterprise plans typically exclude data from model training.

Can the assistant change anything in AROYA? No. The intitial release is strictly read-only.

Can I control which of my users turn it on? Yes, you can manage with members of you facility can access the MCP through the “Manage role” option on the “Settings” >“Staff” page.

Are the numbers the same as in the AROYA app? Yes. The MCP reads from the same database the app uses, so there is no separate copy that can drift. If a figure looks different, it is almost always one of three things: the app rounds or smooths values on charts while the assistant reports raw values; the time window or zone in your question was not the one you had in mind; or the assistant summarized several readings into an average. Ask it to show the exact readings and time range it used and the difference usually resolves.

Can the AI get it wrong? Yes. The AI assistant can misread a question, pick the wrong room, or draw a conclusion the data does not support. Every AROYA MCP conversation carries the reminder "AROYA can make mistakes. Please double-check responses." Treat its answers the way you would a sharp new team member's: useful, fast, and worth verifying before you act on anything important. The app remains the source of truth.

What is the difference between an AI problem and an AROYA problem? Think of the MCP as the assistant that fetches data and the AI model as the one that interprets it. If the assistant returns no data, says it cannot see a facility, or reports a connection error, that is on the AROYA or connector side. If the data is correct but the summary, math, or recommendation is off, that is the AI model's interpretation; rephrase the question or ask it to show its work.

What can it not do today?

• Write or change anything in AROYA (schedules, tasks, setpoints, notes).
• Show photos or images attached to tasks, scouting, or SOPs.
• Answer questions about facilities you do not have access to.
• Replace AROYA alerts or notifications; it answers when asked rather than watching continuously.
• Guarantee response speed on very broad questions (all facilities, all history) — see Troubleshooting.

Will it get better? Yes, on two fronts. AROYA will keep adding tools and data to the connector, and the AI models themselves improve quickly, so the same connection will handle harder questions over time.

Troubleshooting

Still stuck? Contact AROYA Support with: the AI assistant and plan you use (for example "Claude Pro" or "ChatGPT Team"), the facility, the exact question you asked, what the assistant replied, and a screenshot. Connection problems that are not caused by the AROYA feature flag usually sit inside the Claude or ChatGPT settings, and Support can help you tell the two apart.

Tips for better results

Be specific. Name the facility, room or zone, the metric, and the time window. "Average dryback in Flower 2, Zone 3, last 5 days" beats "how is Flower 2 doing."

Start narrow, then widen. Ask about one room first, confirm the answer matches the app, then ask the assistant to repeat the analysis for other rooms.

Ask for the evidence. Phrases like "show the readings you used" or "list the irrigation events behind that total" make the answer easy to verify.

Match the model to the job. Fast models for quick lookups and daily summaries; the most capable model for multi-room comparisons and yield-driver analysis.

Use follow-ups. The assistant remembers the conversation, so you can say "now do the same for Flower 3" or "chart that by day" without restating everything.

Save prompts that work. Keep a note of questions that give you useful answers and reuse them each morning, or turn them into a scheduled report.

Where do I learn more or send feedback? Setup guides and videos are in the AROYA Help Center under Integrations. Send feature requests, favorite use cases, and anything that did not work to your account manager or AROYA Support; what customers ask for shapes what gets added next.

How did we do?