VANTAGE

VANTAGE runs a set of checks across everything a project has uploaded, then writes what it found into the Findings queue with a citation for each one. It is the engine behind the STRUX Workflows instrument and behind Investigate, and it also builds the 3D digital twin.

VANTAGE is a paid per-project add-on at $250/month per project. It is not in the CORE free tier, it is not bundled into the $500 organization fee, and it is enabled one project at a time.

How a project gets VANTAGE

In the project-creation wizard, VANTAGE is on by default. If you leave it on, you are on the paid path: the $500/month organization fee, a one-time $250 setup fee, and the $250/month VANTAGE charge from the first month. Turn it off on the Review step if you want to stay on free CORE. See Creating a project.

To add it to an existing project, open Project Settings on that project. The add-on is per project, so enabling it on one project does not enable it anywhere else.

If VANTAGE is locked, the page tells you which of three things is missing:

What you see	What it means
Subscribe to unlock VANTAGE	VANTAGE is not in CORE. The organization needs the paid path — the $500/month organization fee — before a project can carry the add-on.
AI features are not in your current plan	The subscription is active, but the plan carries no AI at all. VANTAGE is an AI product end to end, so the fix is a plan change.
VANTAGE is not enabled on this project	Subscription and plan are fine; this project does not carry the add-on. An account admin adds it in Project Settings.

The API returns the same answer to anything that tries to start a run without the add-on: "VANTAGE is not enabled on this project. An account admin can add it in Project Settings for $250/mo."

Two kinds of run

Both run on the same engine, produce the same findings, and share the same concurrency budget.

Checklist run. You pick a checklist, and every item in it is executed against the project. The checklist is frozen at the moment you start, so a later edit does not change a run already in flight. This is what the STRUX Workflows instrument starts. See Workflows for browsing, authoring, and starting one.

Investigate run. You type a question in plain language instead of picking a checklist, and the run builds its own check set from the question. Investigate lives on the app home composer: switch the mode control from Auto to Investigate and describe the check. Investigate runs are started from the desktop app. The question is capped at 4,000 characters.

Neither run works on the demo project, which is read only.

The six built-in checklists

These ship with the product and are available on every project with the add-on. Each one runs five checks plus an equipment-discovery pass and a report pass.

Checklist	What it checks	Disciplines
Mechanical / HVAC Review	Equipment schedules against plan callouts, duct sizing against scheduled CFM, hydronic and refrigerant piping completeness, outdoor-air and exhaust rates, and whether every major unit has a sequence of operations	Mechanical, Controls
Electrical Systems Review	Panel schedules against load calculations, one-line diagram consistency, emergency and standby power coverage, mechanical equipment power provisions, and lighting and receptacle coverage	Electrical, Mechanical
Life Safety & Egress Review	Egress paths and exit capacity, fire-rated assemblies, fire alarm and detection coverage, sprinkler coverage, and ADA accessibility	Architectural, Fire Protection, Electrical
Specification vs Drawing Consistency	Each discipline's spec sections against its drawings: mechanical (Div 23), electrical (Div 26), plumbing (Div 22), fire protection (Div 21), and structural (Div 03, 05)	Mechanical, Electrical, Plumbing, Fire Protection, Structural
Structural Coordination Review	MEP penetrations through structure, equipment loading against framing, embeds and anchors, roof structure against rooftop equipment, and below-grade utilities against foundations	Structural, Mechanical, Electrical, Plumbing, Architectural
Cross-Discipline Coordination	Ceiling plenum conflicts, mechanical and electrical room layouts, shaft and riser sizing, conflicting notes across disciplines, and scope boundaries between trades	Mechanical, Electrical, Plumbing, Fire Protection, Structural, Architectural

You can also run your own checklist. A checklist holds up to 50 items and keeps its last 20 revisions.

What a run reads

A run pulls from the project's processed documents, the extracted equipment records, the schedule, and any RFIs and submittals on the project. When a checklist item names disciplines, the run reads up to 100 documents per discipline; an unscoped item reads up to 200 documents across the project.

A document only contributes if it has been processed. A document showing Not processed or Extraction failed on the Documents page contributes nothing to a run. See Processing.

What happens while a run is going

The run starts as soon as you confirm and keeps going server-side, so you can navigate away. Progress plays in a run panel: a progress bar, one row per checklist item, and an Agents on the job rail showing which worker is active.

Phase	What it is doing	Typical time
Planning	Reads the checklist and works out what data it needs	2 to 5 seconds
Gathering	Fetches the documents, equipment, and records it planned for	10 to 20 seconds
Analyzing	Runs the checklist items, 4 to 8 at a time	30 to 60 seconds per item
Synthesizing	Merges related drafts across items so one issue is not reported five times	5 to 10 seconds
Advising	A second, stronger model reviews the HIGH and CRITICAL findings	about 30 seconds
Narrating	Drafts the run's written summary and reviews it	about 15 seconds
Finalizing	Writes the final counts and settles credits at actuals	about 2 seconds

The estimate shown when the run starts is roughly 30 seconds per checklist item.

Each item row shows one of four states:

• Pending. Not started.
• Running. In progress, with the item's current status line underneath. An item that widens its own scope mid-analysis stays in this state.
• Done. Finished. This covers items that passed, items that produced findings, and items that were skipped as not applicable.
• Failed. The item errored, with the reason on the row.

An amber chip on a row is the count of findings that item produced.

The run itself reads as Running, then Complete, Failed, or Cancelled. Cancel stops a run that is still going. If the panel loses contact it says Disconnected and offers Retry; the run may still be processing in the background.

Reading a finished run

The written summary is generated after the findings, and the findings themselves land in the Findings queue for the project. That is where you triage them.

Before findings are published, a reviewer model passes over the HIGH and CRITICAL ones. Findings it suppresses are recorded as dismissed and hidden from the default view rather than deleted, and the run reports how many were suppressed.

A run that finishes with no findings means those checks did not surface anything in the documents it read. Read that as the result of one checklist against one document set, not as a clean bill of health for the project.

Run records are kept for 90 days, then removed.

Limits

Limit	Value
Concurrent runs	2 per person per project, shared between checklist and Investigate runs
Items per checklist	50
Checklist revisions kept	20
Investigate question length	4,000 characters
Documents read per discipline	100
Documents read project-wide, unscoped	200
Run record retention	90 days

Credits

A run places a credit hold when it starts and settles that hold against actual usage when it finishes. You are charged for what the run used, not for the hold. Cancelling a run releases the hold.

Credits are drawn from the project's pool. A project with a zero balance cannot start a run: AI operations return a hard block while the rest of the product stays usable. See Billing and credits.

Troubleshooting

"Maximum 2 concurrent runs per user." You already have two runs queued or running on this project. Wait for one to finish, or cancel it. The limit is per person, so a colleague's runs do not count against yours.

"Project must belong to an organization to run Workflows." The project is not attached to an organization. Runs need one for scoping. Contact an account admin.

The run is stuck at Queued. A run that never starts is reaped automatically after 30 minutes and marked failed with a reason. If that keeps happening, start a fresh run rather than waiting on the old one.

The run finished but read fewer documents than expected. Only processed documents are read. Check the Documents page for anything sitting at Not processed, Processing, or Extraction failed, fix those, and run again. See Processing.

The run is blocked on credits. The project pool is empty. Top up, or wait for the monthly allowance to reset at renewal. See Billing and credits.

Everything VANTAGE is locked. Work through the three gate states in How a project gets VANTAGE in order. Still being on CORE, without the paid path, hides the other two, so fix that first.