Site analysis has always been one of those things that takes longer than it should. You get the survey, the zoning sheet, maybe a Google Earth screenshot, and you spend a few hours pulling it all together before you can sketch a single line. AI site analysis in architecture is starting to change that, not by replacing the judgment call, but by handling the data-reading so you can get to the judgment call faster.

In Snaptrude, two agents handle site work: Analyze Site and Buildable Envelope. This post explains what each one does, how they work together, and where they fit into the way architects actually work.

Why does site analysis slow down early design?

The problem is not that site analysis is hard. It's that it's fragmented.

Your zoning data lives in a municipal PDF. Your survey is a DWG file. Sun path data comes from a separate tool. Setback rules are buried in a footnote. None of it talks to each other, and pulling it into a coherent picture before you can even start designing takes real time.

Here's what we noticed working with architects across firm sizes: the biggest bottleneck in early design is not ideation, it's information assembly. Teams spend hours gathering and cross-referencing data that should take minutes, and by the time they're ready to design, the project timeline has already taken a hit.

The downstream effect is worse. When site reading takes a long time, architects tend to do it once and commit early. You pick a site orientation, establish a setback strategy, and move forward. There's not enough time to test alternatives. That first read becomes the only read.

AI site analysis agents exist to close this gap. They don't replace your judgment about what the site means for the design. They read and process the raw data so your judgment has something to work with immediately.

Key Takeaway: The real cost of slow site analysis isn't the hours spent on it. It's the alternatives you never get to test because you committed too early.

What do the Analyze Site and Buildable Envelope agents actually do?

These two agents work in sequence. Analyze Site reads the site, Buildable Envelope tells you what you can build on it.

Analyze Site

The Analyze Site agent takes your site boundary and location data and produces a spatial reading of the site. This includes sun path and solar exposure across the site at different times of year, prevailing wind direction, topography if elevation data is available, access points from adjacent roads, and any visible constraints like easements or neighbouring structures that affect planning.

What it does not do is tell you what to build. It gives you a structured picture of what you're working with. Think of it as the first page of a proper site analysis document, generated in the time it would take you to open your zoning PDF.

The output is editable. You can add context, flag specific conditions, and build from it. The agent does the first read; you do the interpretation.

Buildable Envelope

Once you have the site analysis, the Buildable Envelope agent takes your zoning parameters, including setbacks, height limits, FAR (floor area ratio), and any overlay zones, and generates the maximum volume you can legally place on the site.

This is the outer boundary of what's possible before design decisions start narrowing it down. In a typical workflow, this is something you'd calculate manually, often across two or three tools. The agent does it in one step and gives you a 3D envelope you can work inside immediately.

The envelope updates when your parameters change. If you're testing two zoning scenarios, or if the brief changes the FAR, you run the agent again and get a new envelope. You're not recalculating by hand every time.

Together, these two agents answer the two questions every architect asks at the start of a project: what is this site, and what can I put on it?

Key Takeaway: Analyze Site reads your site conditions. Buildable Envelope translates your zoning parameters into a 3D volume. Used together, they give you a clear picture of constraints and opportunity before design begins.

Where do these agents fit in your design workflow?

Both agents are useful earliest, before massing begins. That's the right time to use them, and here's why that matters.

If you run Analyze Site after you've already developed a massing concept, you're checking your work rather than informing it. The agent is most useful when it shapes decisions rather than validates them. Run it when the site boundary is confirmed and the brief is in hand, but before anything has been designed.

Buildable Envelope is useful at two distinct points. The first is right after site analysis, to set the outer limits of the design. The second is later in the process, when a design is being pushed against zoning limits and you need a quick reference for where the hard boundary is.

Here's a practical sequence that works for most projects:

1. Import your site boundary or sketch it in Snaptrude.
2. Run Analyze Site. Review the output, add any local knowledge the agent doesn't have, and save it as your site brief.
3. Input your zoning parameters and run Buildable Envelope. You now have a 3D volume to design inside.
4. Move to massing with Snaptrude's Explore Massing agent, working within the envelope the agent generated.

This sequence takes what used to be a half-day of preparation work and compresses it to under an hour. The time doesn't disappear, it shifts. You spend it designing rather than data-gathering.

One thing worth saying clearly: these agents work best when you bring local knowledge they don't have. The agent will read the zoning sheet, but you know that the municipality has been inconsistent about applying a specific overlay, or that the neighbouring building's shadow matters more than the code suggests. That local knowledge still belongs to you. The agent handles the data; you handle the judgment.

Key Takeaway: Run Analyze Site and Buildable Envelope before massing begins, not after. The earlier you use them, the more they shape your design rather than just confirm it.

What should you not expect from these agents?

Honesty about limitations saves time, so here it is.

Analyze Site does not replace a site visit. It reads data. It cannot capture what it feels like to stand on the site, notice that the adjacent building blocks afternoon light in a way the model doesn't fully capture, or pick up on neighbourhood character. Use the agent to process what can be processed, then use your site visit to catch what it missed.

Buildable Envelope depends on accurate zoning inputs. If you give the agent outdated setback rules or miss a height overlay, the envelope it generates will be wrong. Garbage in, garbage out. The agent is only as accurate as the parameters you provide, so verify your zoning data before you run it.

Neither agent makes design decisions. They read constraints and generate envelopes. What you do inside that envelope is entirely up to you. If you're expecting the agents to tell you where to put the building entrance or how to orient the main facade, that's not what they do. That's the design work, and it stays with you.

These are real limits, and worth knowing before you build a workflow around these tools. Within those limits, though, both agents handle a category of work that most architects would rather not be doing manually.

Key Takeaway: These agents read data and generate envelopes. They don't replace site visits, and they don't make design decisions. Know what they do and plan around what they don't.

Start with the site

Every design problem is a site problem first. Before program, before massing, before aesthetics, there is a specific piece of land with specific rules and specific conditions. The clearer your picture of that land at the start, the better the decisions you make inside it.

Analyze Site and Buildable Envelope are built for that first moment. They don't design for you. They clear the path so you can.

If you want to see how these agents work on a real project, try them in Snaptrude or explore all five AI agent groups to see where site analysis fits in the full early-design workflow.

We built these agents because we kept seeing the same thing: architects with strong instincts spending their first day on a project reading PDFs instead of designing. That's the problem we wanted to fix. Whether we got it right is something you'll know within the first hour of using them.