Skip to content
AI at Projexion

Make your project data part of your AI strategy.

AI is only as good as the numbers you feed it. Projexion gives it the cleaned-up, reconciled numbers your portfolio already runs on, then lets you ask in two places: Iris inside Projexion, or Claude and the other AI tools your team already uses, through our MCP server.

Inside Projexion · Iris

Ask Iris, get the chart and the dashboard back.

Iris is the assistant built into Projexion. Type "Which projects are over budget in California?" and you get the answer, a chart, and a link to the dashboard behind it, without leaving the app.

It's the easy starting point for PMs and executives who just want an answer before the Friday WIP meeting.

Charts and short summaries right in the chat
Builds a dashboard when you ask for one
Keeps the thread, so a follow-up builds on the last question
Every conversation saved and searchable
Iris AI assistant answering questions about a construction portfolio in plain English
Your own AI tools · MCP

Already living in Claude? Bring the portfolio with you.

Our MCP server hooks Projexion up to Claude, ChatGPT, Cursor, and any other client that speaks the Model Context Protocol. Your assistant reads dashboards, runs calculations, and finds projects by meaning, all against live data.

That's where it gets interesting. Claude can pull cost-to-complete from Projexion, then draft the owner email or the board memo in the same conversation.

Add to Claude in a couple of clicks, no API key
Works in any MCP client, same tools everywhere
Reads dashboards and runs calculations on live data
Writes back only where your role already allows it
Claude using the Projexion MCP server to summarize portfolio health, with stale data, cost overruns, and schedule slippage by job

Which one should you use?

Most teams use both. Here's the usual split.

Get Projexion into Claude

About two minutes, and nothing to install.

1
Add the connector
Open Projexion in the Claude directory and add it, or paste our server URL as a custom connector in any other MCP client.
2
Sign in to Projexion
A normal browser login. Pick which of your accounts the connection can see, and set a default.
3
Ask
"Which jobs slipped this month, and why?" Claude pulls the real numbers instead of guessing.
Same numbers, same permissions

Whichever AI asks, it sees what you see, and nothing more.

Iris and the MCP server read the same reconciled data that powers your dashboards and Argus briefings, so an answer in Claude matches the number in your WIP report.

Every call checks your own permissions on the server. If you can't see a project or a cost figure in Projexion, neither can the model behind it.

One source of truth behind every answer
Permissions checked on every call
OAuth sign-in, no secrets stored on your machine
SOC 2 Type II compliant

Give your AI numbers it can trust.

Book a demo and we'll show you Iris and Claude answering questions on data like yours.