Tools & Data — Help Center

Practical SOPs for parcel, ordinance, government data, utilities, records, document, image, and web workflows.

What data sources does the agent use?

The agent draws from multiple evidence layers for every project:

• Parcel + ownership context (Regrid, refreshed at bootstrap, with optional user-corrected zoning overrides)
• FEMA NFHL flood polygons — federal hazards.fema.gov primary, Esri Living Atlas fallback
• USDA SSURGO soil polygons — runtime SDA Tabular SQL, with per-parcel polygon-clipped acreage by drainage class
• USGS elevation (3DEP) and seismic (ASCE 7-22) — pre-fetched per parcel during bootstrap
• USFWS NWI wetlands polygons — federal fwsprimary mapper vector service primary, PostGIS self-host fallback for ingested states
• HPMS + Census TIGER roadways — federal HPMS v2 primary with TIGERweb gap-fill for local roads
• OSM roadways — supplementary classification when federal data is sparse
• Reviewed local GIS layers — utilities, jurisdiction boundaries, easements, right-of-way, overlays, rail corridors, watershed, and topography where authoritative public endpoints are available
• County clerk / recorder portals — recorded-plat retrieval where a reviewed public records portal is connected for the county
• Indexed municipal + county ordinance content — Municode, AmLegal, eCode360, plus custom sources
• Your uploaded project documents — searchable corpus per project
• Supplementary web search and research sub-agents — for current agency pages and references not in your project

All data sources are available to every seat — there is no per-tier feature gating. If a result materially affects a design or permitting decision, verify the cited source directly.

What tools can the agent use, and how should I ask for them?

You do not need to name internal tool names, but task-shaped prompts help the agent select the right capability.

Best practices:

• Ask for one evidence family at a time when the decision matters.
• Tell the agent when a source class is mandatory, such as use only uploaded documents and ordinance citations.
• Let the agent ask a follow-up question when a tool returns multiple plausible records.
• Use chat corrections for persistent facts, not just one-off wording fixes.

Email my agent uses the same tools for lookups and deliverables, but disables write-heavy actions (URL ingest, live ordinance scraping, site-plan editing). See Email my agent in Help for routing and billing rules.

What is the SOP for parcel and zoning lookup?

Use parcel lookup early to establish the site facts that the rest of the project depends on.

Recommended workflow:

1. Ask for a parcel summary first: address, parcel identifiers, ownership context, and lot area.
2. Ask for zoning context next: district, overlays, and obvious use or dimensional constraints.
3. If the zoning is wrong, correct it — see "How do I correct a parcel's zoning?" in Chat Workflows.
4. If the site has multiple parcels, ask the agent to compare parcel-specific differences instead of requesting one blended answer immediately.
5. Parcels also carry assessor fields (year_built, building_sqft, assessed_value, improvement_value) that the agent uses automatically to flag pre-1978 lead, pre-1980 asbestos, and high-improvement-value teardown candidates without an extra lookup.

Good prompts:

• Summarize parcel facts and zoning context for this site.
• Compare parcel 1 and parcel 2 for lot size, ownership, and zoning differences.
• What does the assessor data tell us about likely demolition needs on this site?

What is the SOP for ordinance lookup?

Ordinance answers are strongest when you ask for a specific constraint category and require citations.

Best practice:

1. Wait until bootstrap is complete for both municipality and county. The chat-readiness gate holds until both are indexed.
2. Ask one topic at a time: setbacks, parking, landscaping, stormwater, floodplain, or use permissions.
3. Ask for the chapter and section citations, not just a summary.
4. If the municipality uses overlays or special districts, ask for those separately.

Good prompts:

• What code sections govern front, side, and rear setbacks for this parcel?
• Find the stormwater design or drainage requirements that apply to this municipality.
• Show the zoning sections for parking counts and any reduction options.

Avoid broad prompts like tell me about the code because they produce weaker retrieval.

How should I use flood, wetlands, and soil tools before exporting?

Three federal polygon datasets feed parcel-clipped acreage directly into reports and feasibility narratives:

The platform clips each polygon to your parcel boundary, so reports and chat answers reason over actual acreage within your parcel instead of binary "parcel touches Zone AE" signals. You will see phrasing like:

• 0.7 acres of Zone AE within parcel boundary
• 1.2 acres of PEM1A wetland
• 3.4 acres of poorly-drained Drummer silty clay loam

Recommended sequence:

1. Ask for the flood designation, panel context, and any available BFE information.
2. Ask for wetlands proximity and mapped wetland classes near the parcel.
3. Ask for soils breakdown by drainage class and hydric percentage.
4. Review the cited findings — open the parcel-map view modes if you want to see the polygons painted directly.
5. Request a flood map, wetland map, or combined site-hazard map export if you need a deliverable or appendix item.

This avoids creating polished exports from unverified assumptions.

How do I use the project map modes?

The Project Home map uses a layer dropdown so you can inspect spatial evidence without leaving the project. Modes may appear even when disabled or empty, which helps you see what data class was checked.

Layer groups in the dropdown mirror the workflow categories: Parcel & Boundary, Hydrology & Water, Terrain & Geotechnical, Infrastructure & Access, and Environmental & Subsurface.

Tips:

• The Flood, Wetlands, Soil, Roads, Utilities, and Topography views show the same evidence families the agent uses in chat and reports.
• In Regulatory & Site Constraints, the map opens on mapped easements that affect your site. Overlapping mapped easement records are consolidated into inspectable corridors with a count badge, and you can expand any corridor to see the individual source records. Nearby context and planning (zoning) overlays are independent toggles. These are public GIS screening layers — not a title commitment — so verify against a recorded survey.
• A disabled or warning state usually means the project has no data for that layer yet, bootstrap is still running, or the jurisdiction does not publish a reviewed public GIS endpoint.
• Use the LifeBuoy button if a public layer should exist but the map cannot find it.
• Topography uses labeled contour data where available; treat it as planning context unless your uploaded survey supersedes it.

How do I export GIS data for CAD, QGIS, or Google Earth?

Use Export GIS Data from the Project Home map when you need layers outside PrecisionDocs.

Available layers depend on the project and jurisdiction, but the export dialog can include:

• Project Parcels
• Neighbor Parcels
• FEMA Flood Zones
• USFWS Wetlands
• USDA Soils
• Roads
• Utilities
• Rail Corridors
• Watershed
• Jurisdiction Boundaries
• Easements
• Right-of-Way
• Zoning Overlays

AutoCAD note: DXF is the CAD hand-off format. DWG is not supported from the Project Home GIS export.

Best practice: export only after the map layer has loaded cleanly and warnings have been reviewed. If a layer is omitted, the export should be treated as a data-availability gap, not proof that the feature does not exist.

How can the agent retrieve recorded plats and county records?

Ask for a recorded plat when you need the legal subdivision exhibit, county-record citation, or a handoff link to the official records portal.

How the workflow behaves:

1. The agent resolves the subdivision identity from project GIS layers, parcel legal description, and owner context.
2. It searches a reviewed county clerk or recorder portal when one is connected for the county.
3. If the match is high-confidence and corroborated by a document identifier, the agent can retrieve and ingest the public document link when available.
4. If there are multiple plausible matches, the agent asks you to choose instead of guessing.
5. If the portal requires login, payment, or an account, the agent hands back the free index citation and official deep link. It never bypasses login or payment.

Good prompts:

• Retrieve the recorded subdivision plat for this project and cite the county record.
• Find the recorded plat or explain what county record information is missing.
• Use the parcel legal description to look for the subdivision plat.

If the agent cannot retrieve the plat, ask for the best available handback: document number, book/page, subdivision name, recorder portal link, and the reason retrieval stopped.

How does the agent know about neighboring parcels and roads?

Bootstrap runs a spatial awareness pass that classifies adjacent parcels and roads relative to each project parcel.

Parcels:

• Direct — touching the project parcel boundary
• Across-street — separated by a road right-of-way (inferred from negative-space gap when DOT data is sparse)
• Nearby — within about 650 ft to 1.2 miles, used for context but not adjacency

Each neighbor is tagged with which of your project parcels it abuts (source_parcel_indices), so the agent can talk about "the parcel directly south of parcel 2" without you having to point.

Roads:

• HPMS v2 is primary (federal Federal_User_Community service) — gives you AADT, functional class, ownership for arterials and collectors.
• Census TIGER Layer 8 fills HPMS gaps — captures local subdivision streets that HPMS does not include.
• OSM is supplementary classification.

Roads are clipped to per-parcel frontage segments so you get frontage length per parcel, broken out by tier and ownership. Visible in the parcel map's Roads mode.

What does the agent know about my parcel's structures?

During bootstrap, two complementary sources populate structure context:

1. Assessor fields from Regrid are loaded into every parcel: year_built, building_sqft, assessed_value, land_value, improvement_value. The agent reads these from the stable system prompt without an extra tool call, so it can flag pre-1978 lead, pre-1980 asbestos, and high-improvement-value teardown candidates instantly.
2. Satellite vision detection runs Claude Sonnet over ArcGIS World Imagery and enumerates building-like visible structures (buildings, sheds, silos, tanks) as structured JSON. Pavement and vegetation are tracked as site signals, and imagery provenance is recorded when ArcGIS metadata is available. Empty structure results mean the parcel reads as vacant unless assessor evidence says otherwise.

These feed two automatic considerations in the Feasibility Assessment:

• Demolition — fires once on the first parcel meeting any of year_built < 1980, building_sqft > 5000, or a high improvement-value threshold.
• Site inspection — fires when Vision detected at least one visible structure (so you know the report assumed there is something to walk around).

You can ask the agent things like:

• What does the assessor say about the structures on this site?
• Are there any visible structures from satellite imagery?
• Should we be planning for demolition based on the assessor data?

What is the best workflow for document search and image review?

Treat uploaded documents as a searchable project library, not just attachments.

Recommended SOP:

1. Upload the source documents you actually want cited.
2. Wait for the file chip to reach Ready.
3. Ask the agent to search by topic, filename, or report type.
4. If the answer references an image or figure, use the image workflow for more detail.

Examples:

• Search the uploaded geotech report for groundwater depth.
• What does the uploaded zoning letter say about allowable use?
• Describe the uploaded site exhibit image and call out anything important for setbacks or access.

Image-focused review is most useful when a figure, map, or scanned note carries information that a text summary might miss.

When should I use web search versus ordinance or document search?

Use the project knowledge base first. Use catalog GIS/API tools for project-specific physical evidence, contextual web modes for current situational context, and generic web search only when neither narrower path fits.

Use ordinance search when you need municipal code language.

Use document search when the answer should come from your uploaded files.

Use catalog GIS/API tools for project-specific physical evidence such as easements, right-of-way, overlays, utilities, roads, flood, wetlands, soils, and recorded subdivision context.

Use contextual modes of web search for current/situational items such as:

• Active permits and permit applications
• Planned developments and planning agendas
• DOT projects, road work, and capital improvements
• Current agency pages and provider notices

Use generic web search only when neither the project knowledge base nor a narrower catalog tool or contextual web intent fits, or when reading a specific URL.

Web search is helpful, but it is the least authoritative layer if an official ordinance, agency dataset, or project document already exists.

Other help topics

• Getting Started
• Chat Workflows
• Email my agent
• API & connectors
• Collaboration
• Bootstrap & Ordinances
• Documents & Reports
• Credits & Usage
• Billing & Subscriptions
• Account & Settings
• Troubleshooting

• All help topics