Canvas
courses & learning
Learning Enterprise · AI Enablement
One governed door to every system of record.
Our teams live in a dozen different tools. The MCP Gateway is the single, controlled door an AI assistant walks through to reach all of them — so a person can ask a plain-language question and get a real answer, without ever holding a key, touching a database, or seeing something they shouldn't.
One front door
Scoped per person
Redacted by default
Every call logged
The shape of it · one picture
A controlled path between people and the records.
Why a gateway at all
The problem it quietly solves.
Give a capable assistant to a whole organization and the same three problems appear every time. People paste sensitive records into a chat box to get work done. Everyone needs credentials to a dozen systems just to answer one question. And nobody can see, afterward, what was actually asked or answered.
The Gateway turns all three into a single, well-lit hallway. The assistant never gets raw keys — it gets a door. What comes back through it is already scoped to who you are and already stripped of anything private. And every trip leaves a footprint you can review later.
What every request gets — asked for or not
The four promises the door makes.
01
Least privilege, per person
Access is granted by role, not by handing out passwords. A marketing analyst and a learning designer walk through the same door and reach different rooms. You can't accidentally use a source you were never given.
02
Privacy is the default, not a setting
Personal and student information is stripped out before an answer ever reaches the assistant — a built-in requirement, not something a user has to remember. The rare, justified case that genuinely needs it takes a separate, explicit path.
03
Everything is on the record
Every request is counted and traceable — who asked, of what source, and whether it was allowed. That's what turns "an AI touched our data" from a worry into something you can actually audit.
04
Shared know-how travels with it
Alongside the data doors sits a shelf of reusable skills — captured expertise for common tasks, so a good workflow one team figures out becomes something the whole organization can reuse instead of reinventing.
The systems people reach — named, not the plumbing
What sits behind the door.
Each system gets its own connector: a small adapter that knows how to speak that tool's language, so the person asking doesn't have to. New systems join by adding a connector, not by rewiring anything.
Salesforce
the learner & partner record
HubSpot
outreach & lifecycle
Metabase
dashboards & questions
BigQuery
the warehouse
Microsoft Clarity
behavior & friction
Notion
docs & knowledge
Stripe
payment records
Sentry
product health
Airtable
team workbooks
Ahrefs
search & web reach
ZoomInfo
market & account intel
The list grows. The door doesn't change.
What it is — and firmly is not
Drawing the line.
■ It is
- A single, governed entry point to many systems
- A scope check that reflects who you are
- A privacy filter that runs before you see anything
- An audit trail for every request
- A shared shelf of reusable skills
○ It is not
- A pile of raw credentials handed to a chatbot
- A way to route private data to a model
- A place to store learner or partner records
- An unaccountable back channel to production
- Something a person has to secure by hand
The point of all of it
Who it's for.
It's for the person in operations who wants a number without learning SQL. For the designer who wants to know where learners get stuck without exporting a spreadsheet of names. For the analyst chasing a question across four tools that were never meant to talk to each other.
The promise is small and specific: ask in plain language, get a real answer, and trust that the guardrails held — because they were built into the doorway, not bolted onto the person.