Skip to content
A tour of the app with example data: an answer about the next meeting streams in, each call shows its latency, tokens and estimated Copilot Credits, then the code, the three routes and what each route can read and change.

Work IQ integration lab

Three ways to build Work IQ into your own tools, MCP, A2A and the REST API, each as a live chat next to the code that runs it.

Live chats over your data

Each route is a real chat over Microsoft 365 data, read with the signed-in user’s permissions. There are no sample answers.

Every call in view

Flow shows each step between you, the model and Work IQ, with latency, tokens, a Copilot Credits estimate and the raw payloads.

The code that runs it

Code shows the app’s actual source beside the chat: a Deep Agents harness, A2A JSON-RPC and two REST calls.

Three routes, side by side

MCP, local or remote, with your own model; A2A to delegate a task; REST for a grounded chat inside your app.

Sign-in, explained

What each route needs to sign in, what an admin grants once, and the recommended setup for your own tool.

Windows, macOS and Linux

Download the app for x64 or Arm, or run it from source with Node.js.

Every route needs a Microsoft 365 tenant with Work IQ enabled and a usage-based billing plan; both MCP routes also need a model with tool calling. See What you need.

The illustrations are AI-generated (MAI-Image-2.6). The screenshots and the demo show example data.