AI-native APS: From idea to outcome, with context, control, and trust

This blog may contain forward-looking statements about our products and product capabilities, and strategies. Statements regarding planned or future development efforts for our products merely reflect our current plans and are based on factors currently known to us. Purchasing decisions should not be made based upon reliance on these statements.

At this year's Platform Leadership Forum at AU, we shared how Autodesk Platform Services is becoming AI-native – working with you at every stage of your development journey. If you missed it, that recap is here: The new APS: A developer platform for the AI era.

In this post, we will walk through what AI-native APS looks like in practice by following one developer, and one app, from a problem statement to a distributed product.

The app: a Revit model viewer that extracts element parameters and schedules for filtering. Simple to state. As you'll see, harder to get right, and that's exactly where an AI-native platform changes the developer journey.

Build with context from the start

Every developer starts with the question: can APS solve this, and what's it worth? Since June, you've been able to ask Autodesk Assistant on APS directly, and it solves two problems developers used to have to answer on their own - what's the right technical approach, and what's the business case for building it.

Here's the prompt our developer gives Assistant:

"I want a Revit model viewer that extracts all element parameters and schedules and filters them. What's a simple and cost-optimal way to build this app?"

Assistant recommends the Model Derivative API paired with APS Viewer, and grounds that recommendation in cost. For development, the APS Free Tier covers most testing. For production, Viewer is free to use, and it lays out what the Model Derivative API usage will run at scale.

It also flags something a developer might not catch until much later: Revit schedules, which are the organized tables that pull together parameters across multiple elements, aren't extracted directly by the Model Derivative API. To preserve them, you'd need to either extract individual element properties and reconstruct schedule-like data yourself, or use the Revit Design Automation API to export schedule tables separately. That's the kind of gap that normally surfaces after you've already built the wrong thing. Here, it surfaces before you write a line of code.

This is what building with context means: you're not starting from a blank page or a stack of API docs. You're starting from a grounded recommendation, the trade-offs already named, and a real number to weigh it against.

First build done with more room to fine-tune

Our developer decides the schedules matter too much to drop. That changes the build. Now it needs Revit Design Automation API in the mix alongside Model Derivative API and the Viewer. Before, this is where the real time cost would show up: researching Revit Design Automation's workflow, figuring out how it sequences with the APIs already in play, wiring up authentication, and getting all of it working together.

With the APS Workflow MCP that first build compresses from a research project into a conversation. Here's the prompt:

"I need to build a Revit model viewer and parameter extraction app. It should load a Revit model into the APS Viewer and simultaneously extract all element parameters and schedules via Revit Design Automation. Use the APS Workflow MCP — keep it composable."

The agent understands the dependencies across all APIs involved, sequences them correctly, and hands back a working, composable app in minutes. The developer’s work moves up the stack – with more time for reviewing what got built, refining it, and deciding where to take it next.

APS Workflow MCP is in development, and this is a good time to help shape it: sign up for the private beta.

Manage with transparency and control

The app is live. Now a different set of questions shows up – not "can I build this," but "how is it actually being used." When is it busiest? How many tokens is it consuming, month to month? If more users get added, how does that usage curve change? And critically, this isn't about visibility into one app, but across your whole portfolio.

While that data exists today, it's hard to find across dashboards and reports you have to search. We will soon bring usage insights directly into Autodesk Assistant, so you can just ask it – the same way you asked it to scope the build in the first place. Once the insights are readily visible, you can act on them. Retire apps nobody's using. Take action before a team exceeds its token allocation.

Distribute with trust

Our developer was able to refine their app with specific functionality thanks to the time saved in initial development. Now they want take it beyond their organization and open it to the industry.

That's where being verified and published on the Design and Make Marketplace comes in.

A verified app becomes available to Autodesk Assistant – today, if a customer goes to the Design and Make Marketplace and asks for a model-viewing solution with this kind of functionality, Assistant can surface it directly. That's real distribution. But it still depends on the customer coming to Marketplace and knowing to ask.

What's coming next closes that gap: Assistant inside Autodesk products will find verified apps and connect to verified MCPs, recognizing the workflow gap in the moment and completing the outcome defined by the user. This is the difference between being discoverable and being present at the point of need, and trust through verification makes it possible.

That's AI-native APS with compounding advantage across the life of what you build – from the start of an idea to reaching users and agents with your solution when they need it.

If you're working on a solution idea, or don't yet know where to start with APS, explore your problem space with Autodesk Assistant. Want early access to the faster build we walked through here? Sign up for the APS Workflow MCP private beta. And when your solution is ready for market, get it verified and published on the Design and Make Marketplace.