Ask Your Facility Anything: Introducing the Elevated Signals MCP Server
Connect AI assistants like Claude and ChatGPT directly to your Elevated Signals data — securely, in plain English, with no risk to the integrity of your records.
If you've used an AI assistant like Claude or ChatGPT, you already know what they're good at: answering questions, summarizing documents, drafting reports, spotting patterns. And you've probably also noticed their biggest limitation — they don't know anything about your operation. Ask ChatGPT how many batches you harvested last month and it can't tell you, because it has no connection to your data.
That changes today. We've released the Elevated Signals MCP server: a secure bridge that lets the AI assistant you already use read your live Elevated Signals data and answer questions about it directly.
Instead of exporting a spreadsheet, filtering it, and building a pivot table, you can now ask:
"Which batches are currently in the drying stage?"
"Give me a plain-language summary of this batch's recent activity."
"What inventory is at my packaging room right now?"
…and get an answer in seconds, grounded in your actual records.
This guide explains what's behind that capability, how to set it up, and — just as importantly — how to use AI well in a regulated manufacturing environment. Whether your team is already deep into AI or hasn't touched it yet, this is the foundation for everything we're building next.
A quick primer: what AI assistants are (and aren't)
If your team is new to this, here's the honest version — no hype.
Modern AI assistants are built on large language models (LLMs). They are genuinely excellent at a specific set of tasks:
Answering questions in plain language. No query syntax or report builder. You ask the way you'd ask a colleague.
Synthesis and summarization. Condensing hundreds of records into a readable summary, or pulling the thread that connects them.
Drafting. Turning raw data into a status update, a meeting brief, or the first draft of a report.
Pattern-spotting. Noticing that one room, one supplier, or one SKU keeps showing up in your deviations.
They also have well-known weaknesses, and in a regulated environment you should know them cold:
Without access to your data, they guess. An AI with no connection to your systems will sometimes produce confident-sounding answers that are simply made up (the industry calls this hallucination). Connecting the AI to your real data — which is exactly what the MCP server does — is the single most effective way to ground its answers in fact.
They are assistants, not systems of record. An AI summary is a starting point, not a quality record. Your batch records, audit trails, and electronic signatures live in Elevated Signals, where they're governed, versioned, and inspection-ready. The AI reads from that system; it doesn't replace it.
Outputs should be verified before they matter. For anything that feeds a regulatory decision, treat AI output the way you'd treat a junior analyst's work: useful, fast, and reviewed before it goes in the file.
This is exactly why our first release is read-only — more on that below.
What is MCP?
MCP (Model Context Protocol) is an open standard, originally created by Anthropic (the makers of Claude) and now adopted across the industry — including by OpenAI (ChatGPT), Google, and Microsoft. It defines a common way for AI assistants to connect to external tools and data sources.
The easiest analogy: MCP is like USB-C for AI. Before USB-C, every device needed its own proprietary cable. Before MCP, every AI integration needed its own custom plumbing. With MCP, any compliant AI assistant can plug into any compliant data source — one standard connector, many devices.
The Elevated Signals MCP server is our end of that connector. It runs as part of your existing Elevated Signals environment and exposes your operational data — batches, inventory, quality events, and more — in a form AI assistants understand. You don't install anything on your servers; it's already there.
Read-only by design
Version 1 of the MCP server is read-only, and that's deliberate.
What read-only means in practice:
The AI can look things up: query records, retrieve details, aggregate, summarize, and analyze.
The AI cannot change anything: it can't create, edit, sign, or delete records, and it can't trigger workflows. Your audit trail, your data integrity controls, and your validated processes are untouched.
Access is tied to your account and your permissions. The AI sees what you're allowed to see — nothing more.
We think this is the right way to introduce AI into a GMP environment: observe first, act later. You get the productivity gains of natural-language access to your data, with no risk to the integrity of the records your licenses depend on.
Read/write capabilities — AI that can draft records, create tasks, or update inventory on your behalf — are on our roadmap and available for early access today. Because writing to a system of record carries real compliance implications, write access is enabled per-customer in partnership with our team rather than switched on by default. If you're interested, reach out and we'll walk you through it.
What happens to your data
For a regulated operation this is the question that matters most, so here it is plainly:
Nothing leaves Elevated Signals on its own. The MCP server is read-only and passive. Data is only ever retrieved in response to a question you ask through your connected AI assistant — there is no background sync, export, or copy.
When you ask a question, the relevant records are sent to your AI provider (for example Anthropic or OpenAI) to generate the answer — exactly as they would be if you pasted them into a chat yourself. Elevated Signals is the data source; the AI provider does the reasoning.
How that data is handled, stored, and whether it is used for model training is governed by your own agreement with the AI provider — not by Elevated Signals. On most Team, Business, and Enterprise plans, the major providers do not use your data to train their models by default. Confirm the specifics under your organization's plan and Data Processing Agreement before connecting production data.
Access is scoped to your permissions. The connector authenticates as you, so the AI can only read what your Elevated Signals account is already allowed to see.
If you have specific data-residency, retention, or DPA requirements, talk to us and your AI provider before connecting — we're happy to help you get the right setup in place.
Getting connected
You'll need two things:
Your MCP server URL. It follows this pattern, using the same subdomain you use to log in to Elevated Signals:
So if your team logs in at acme.elevatedsignals.com, your MCP URL is https://acme.elevatedsignals.com/api/mcp.
2. An AI assistant that supports MCP. Below are setup steps for the two most popular options. The same server works with any MCP-compatible client, including Claude Code, Microsoft Copilot Studio, and a growing list of others.
Connecting Claude (recommended)
Custom connectors are available on Claude Pro, Max, Team, and Enterprise plans. The connection works the same way across claude.ai, Claude Desktop, Cowork, and the mobile apps.
How you connect depends on your plan. Individual accounts add the connector themselves. On Team and Enterprise plans there are two roles: an admin adds the connector once for the whole organization, and then each team member enables and signs in to it individually.
If you're on Pro or Max (individual account):
Go to Customize → Connectors, click the + button, and select Add custom connector. (Connectors recently moved here from Settings — if an older guide says "Settings → Connectors," that page now just points you to Customize.)
Give it a name (e.g. Elevated Signals) and paste your MCP URL.
Click Add, then sign in with your Elevated Signals credentials when prompted to authorize the connection.
Open a new chat, open the tools/connectors menu, and make sure the Elevated Signals connector is enabled.
If you're a Claude Team or Enterprise admin (Owner) — set it up for your organization:
Go to Organization settings → Connectors. (Only an Owner or Primary Owner can add a connector for the organization.)
Select Add custom connector, give it the name Elevated Signals, and paste your MCP URL.
Click Add. This makes the connector available to everyone in your organization — but it does not sign anyone in. Each member still authenticates with their own Elevated Signals credentials, so Claude only ever accesses the data that user is permitted to see.
Let your team know it's available and point them at the steps below.
If you're a regular Claude Team or Enterprise user — connect the one your admin added:
Go to Customize → Connectors. You should see Elevated Signals already listed (added by your admin). If you don't see it, ask your workspace Owner to add it first.
Click Connect next to it and sign in with your own Elevated Signals credentials when prompted.
Open a new chat, open the tools/connectors menu, and make sure the Elevated Signals connector is enabled.
Verify it's working: ask something simple like "What tools do you have available from Elevated Signals?" or "List my most recent batches." You should see Claude call the Elevated Signals connector and return real data from your environment.
Connecting ChatGPT
OpenAI now calls these integrations apps (formerly "connectors" — the rename happened in December 2025), and custom MCP apps currently require Developer Mode, which is in beta:
Individual (Plus/Pro) accounts: Developer Mode is available to ChatGPT Plus and Pro subscribers. Enable it in Settings, then add a new app/connector with your Elevated Signals MCP URL.
Business / Enterprise workspaces: a workspace admin must enable Developer Mode (under Workspace Settings → Permissions & Roles → Connected Data), after which admins can add, test, and publish the app for the workspace.
Once added, enable the Elevated Signals app in your chat's connected-apps menu and run the same verification prompt as above.
Connecting Microsoft Copilot
Microsoft Copilot supports MCP, but — unlike Claude and ChatGPT — it does not let an individual user add a custom connector from inside the Copilot chat window. Connecting Elevated Signals to Copilot is done by a maker or administrator, through one of two routes:
Copilot Studio (recommended): In your agent, go to Tools → Add a tool → New tool → Model Context Protocol, then enter a name, description, and your Elevated Signals MCP URL, and set the authentication type. Generative orchestration must be turned on. Once you publish the agent, your team can use it from Microsoft 365 Copilot.
Microsoft 365 admin center: An administrator can add Elevated Signals as a custom (federated) connector under Copilot → Connectors → Gallery, then make it available to users. Invocation from Copilot Chat and Excel can be inconsistent today, so Copilot Studio is the more reliable route.
If you're an individual user without Copilot Studio or admin access, there is no "add custom connector" option inside the Copilot chat itself — the quickest path is to connect through Claude or ChatGPT instead (see above). Need a hand with a Copilot Studio setup? Email help@elevatedsignals.com.
Other tools
Because MCP is an open standard, the same URL works anywhere MCP is supported — developer tools like Claude Code and Cursor, agent platforms, and enterprise AI gateways. If your team uses something we haven't listed and you'd like help connecting it, get in touch.
What can you do with it?
Here's where it gets practical. These are the kinds of questions early users are asking — use them as starting points and adapt them to your operation:
Production visibility
"List my most recent batches."
"How many plants are in each growth stage right now?"
Quality and compliance
"Pull up the records logged against a specific batch."
"Show me the details and audit history of a specific record."
Inventory
"What inventory is at a specific location?"
"Show me the inventory for a specific batch."
Reporting and prep
"Draft a short status update on a batch I can share with my team."
"Summarize a specific work order — what went in and what came out."
A few tips for getting good results:
Be specific about time ranges and scope. "Last 30 days" beats "recently."
Ask follow-ups. The real power is conversational: get a summary, then drill into the outlier.
Ask the AI to show its work. Prompts like "list the specific records you used" make verification fast.
Verify before it matters. For anything feeding a compliance decision, click through to the underlying records in Elevated Signals. The AI gets you to the answer faster; your system of record is still the authority.
Troubleshooting
The connector doesn't appear after you add it: connectors are loaded when a chat starts, so one you just added usually won't show up in a conversation that's already open. Start a new chat and check the tools/connectors menu again.
Connection fails when adding the connector: double-check the URL — it must start with https://, use your subdomain, and end in /api/mcp exactly.
Authentication errors: sign out and reconnect the connector. If it persists, your session or credentials may have expired — re-authorize from the connector settings.
The AI says it can't find data you know exists: check that the connector is actually enabled in the current chat, and remember the AI only sees data your Elevated Signals account has permission to view.
Still stuck? Contact us at help@elevatedsignals.com and we'll get you sorted.
What's next
This release is the foundation. Read-only access already changes how quickly your team can answer questions — but it's step one of a longer roadmap:
Read/write agents that can draft records, log events, and update inventory under your review (available now via early access, as described above).
Multi-step agentic workflows — AI that doesn't just answer a question, but completes a process: investigating a deviation, assembling a batch summary, preparing inspection documentation.
Deeper analysis across more of your operational data.
Regulated manufacturing has historically been the last industry to get good software. We don't think it should be last to get good AI — and we intend to make sure Elevated Signals customers are first in line, with the compliance guardrails this industry demands built in from day one.
If you connect your assistant this week and find something useful — or something confusing — tell us. The next version is being shaped by exactly that feedback.
Questions? Reach out to help@elevatedsignals.com or your customer success contact. We're excited to roll this out and can't wait to see what you ask first.