AI workflow

Claude for Commercial Real Estate: What It Does Well and the Data It Needs

What Claude is good at for CRE teams, how Projects, Skills, connectors, Claude Code and the API fit together as of November 2026, and why any answer about a specific property needs a parcel data connection and a citation.

By Nathan Robinson, Founder, Plotzy Published 14 min read

Claude for commercial real estate means using Anthropic's Claude models to read ordinances, leases and offering memoranda, extract data, draft documents and run analysis. Teams use it through the Claude apps, Claude Code or the API. Claude has no parcel, zoning or ownership records of its own. Any answer about a specific property needs a connection to that data and a citation back to the source.

Product names, plans and prices on this page were checked on Anthropic's and OpenAI's own sites on November 19, 2026. Both companies change plans often, so confirm before you buy.

What is Claude good at for commercial real estate work?

Claude is strongest on long documents, structured extraction, drafting in a set format, and analysis it can check by running code. Those four skills cover most of the reading and writing in acquisitions, entitlements and brokerage.

  • Reading long documents. Current Claude models have a 1M-token context window, which Anthropic puts at roughly 555,000 words. A district's zoning sections, a 90-page lease and a planning staff report fit in one request. The API accepts up to 600 PDF pages per request, and the Claude apps accept PDFs up to 1,000 pages.
  • Structured extraction. Structured outputs return JSON that matches a schema you define, such as one row per lease with rent, term, options and exclusives. The Citations feature returns the exact passages that support each claim, so a reviewer can check the source line.
  • Drafting. Letters of intent, investment committee memos, owner outreach letters and zoning memos, written in your template. Claude for Word edits with tracked changes; a Skill can hold your house format.
  • Analysis with code. Claude can write and run Python on a parcel export, a rent roll or a county GIS layer. Claude for Excel builds and updates models in place.

Where a person stays in the loop: any legal conclusion, any planned development (PD) condition, and any fact a deal depends on. Treat Claude's answer as a cited first draft, then confirm with the jurisdiction. Our guide to finding out what a property is zoned for covers the official sources.

What are the Claude products, and which one fits a CRE team?

Claude comes as chat apps, a coding agent and a developer API, with Projects, Skills and connectors layered on top. Most CRE teams start in the apps and add connectors or the API when the same task repeats across many parcels.

SurfaceWhat it isPlans (Nov 2026)Typical CRE use
Claude appsChat on web, desktop and mobile, with web search, file uploads, code execution and ArtifactsFree and upOne-off questions on an uploaded lease or code chapter
ProjectsA workspace with its own chat history, knowledge base and instructionsFree (up to 5 projects) and up; shared projects on Team and EnterpriseOne project per deal or market, holding the code chapters and your buy box
SkillsFolders of instructions, scripts and resources that Claude loads when a task calls for themAll plans; requires code executionYour zoning memo format, LOI template, or buy box rules
Connectors (MCP)Links to outside tools and data through Model Context Protocol serversAll plans; Free is limited to one custom connectorYour CRM, data room, or a parcel data API
CoworkClaude working across your files and tools, including scheduled tasks and pluginsPaid plans, rolling out; now part of ClaudeA weekly agenda scan or a folder of leases reviewed overnight
Claude for Microsoft 365 and Claude in ChromeClaude inside Excel, PowerPoint, Word and Outlook, and in the browserPaid plansUnderwriting models, IC decks, outreach drafts
Claude CodeAn agentic coding tool for the terminal, IDE, desktop app and webPro and upScripts that pull county GIS data and build pipelines
Claude APIThe models, billed per token, with batch, caching, citations and structured outputsPay as you goRunning a check on thousands of parcels

Plan prices on November 19, 2026: Free $0; Pro $20 a month, or $17 a month billed annually; Max $100; Ultra $200. Team seats cost $25 a month ($20 billed annually), or $125 ($100) for premium seats with 5x the usage. Enterprise is $20 a seat a month billed annually, with a 20-seat minimum, plus usage at API rates.

Skills for real estate. A Skill is a folder with a SKILL.md file of instructions plus optional scripts. A CRE team might write one that holds its buy box and the questions to ask of every use table. A script inside it can run the floor area ratio and units per acre arithmetic. The Skill format is published as an open standard at agentskills.io.

MCP for real estate. The Model Context Protocol (MCP) is an open standard for connecting AI applications to outside data and tools. A custom connector points Claude at a remote MCP server. The server must be reachable over the public internet on port 443, or allowlist Anthropic's IP ranges if it sits on a private network. Some parcel data vendors now offer MCP servers for their data. Check what the data license allows before records flow into a model.

What can't Claude do on its own?

Claude cannot tell you what a parcel is zoned, who owns it, or what the code allows there, unless you give it that data. The gap shows up in four ways.

  1. No property records. Current models have a reliable knowledge cutoff of June 2026. They do not hold county assessor, zoning map or deed data for your parcels.
  2. Official sources resist fetching. Web search finds pages, but zoning lives in GIS viewers and code libraries built for browsers. Scripted requests to Municode's code library in November 2026 returned a bot check instead of code text.
  3. Codes change. Ordinances are amended through the year, overlays add rules, and a planned development is governed by its own adopting ordinance and site plan. A model's memory of a code is a snapshot at best.
  4. Addresses mislead. A mailing city does not tell you which government zones a parcel. In Hillsborough County, many parcels with a Tampa address sit outside the city limits.

There are three ways to close the gap. Upload the records for a one-off question. Connect a data source through MCP for live lookups. Or build a pipeline that feeds parcel, zoning and owner data to Claude at scale, with every answer citing its source.

Seven CRE workflows: what Claude does and what data it needs

Each workflow below pairs a task Claude handles well with the data it cannot supply itself.

WorkflowWhat Claude doesData it needsHuman check
Zoning question on one parcelReads the district's use table, dimensional standards and overlays, and answers with section citationsParcel ID, the official zoning and future land use designation, the current code text, any PD ordinanceConfirm with the jurisdiction or a zoning verification letter
Buy-box screen across a countyApplies your criteria to every candidate parcel and ranks the results, with reasonsBulk parcel records with zoning, acreage and land use codes, plus the code text for each districtReview the top of the list and every PD parcel
Ordinance comparison across citiesAnswers the same questions (by-right uses, height, parking) for each city in one tableEach city's current code, with the amendment dateSpot-check cited sections
LOI or OM first draftDrafts in your template from structured deal factsParcel record, zoning, sale history, rent roll, your deal termsA person signs off on every number
Owner and entity researchTraces an LLC owner of record to its managers and registered agentOwner of record from the property appraiser, Secretary of State filingsVerify before outreach; see is the owner of an LLC public record
Planning agenda monitoringReads new agendas on a schedule and flags rezonings near your targetsAgenda and staff report feeds from each city and county, your target listRead the flagged staff reports
Rent roll and lease extractionTurns rent rolls and leases into rows for an underwriting modelThe rent roll and leases, your model templateTie totals to the source

Three notes. Ordinance comparison is where the long context window pays off, so ask for the section number and amendment date in every cell. Owner research touches personal data: keep to public records and log the source of every name (see finding commercial property owners). Agenda monitoring suits scheduled tasks, but each city posts agendas its own way. For the zoning, screening and diligence workflows in depth, see AI zoning analysis, AI site selection and AI due diligence.

Built for your team

Have Plotzy build this workflow on your criteria

Plotzy builds AI workflows that run on parcel, zoning and owner data: your buy box, your markets, your review steps, with every answer traced to its source. Start with a 30-minute call about the workflow you want automated.

  • Working prototype in the first 2 weeks
  • Answers cite the code section, record or dataset they came from
  • Outputs land in your CRM, spreadsheet or deal pipeline
Book a 30-minute call

30 minutes, no pitch deck

Worked example: a buy-box screen in Hillsborough County, Florida

A land team wants every vacant commercial or industrial parcel in Hillsborough County checked against its criteria. The Hillsborough County Property Appraiser publishes a daily parcel layer on the county's ArcGIS Online site, last edited November 18, 2026. Claude Code can write the queries; the counts below come from that layer.

StepInput or queryResult
1. Count all parcelsProperty Appraiser daily parcel layer531,626 parcels
2. Filter to vacant landUse code 1000 (vacant commercial) or 4000 (vacant industrial)2,808 parcels, about 5,456 acres
3. Group by zoning codeThe layer's zoning field83 distinct codes. Largest: PD 677, CG 431, CI 348, IG 240, M 219, CN 130
4. Flag planned developmentsCodes containing "PD"742 parcels whose rules sit in a site-specific ordinance
5. Check the rest against the codeClaude reads each district's sections once and applies them to every parcel in itA ranked list, each row citing the code section
6. Route exceptions to a personPD parcels, split zoning, overlay hitsA review queue

What the example shows:

  • Most of the work is data, not the model. The parcels fall under four zoning authorities, each with its own code: the City of Tampa, Plant City, Temple Terrace and unincorporated Hillsborough County. See our guides to Tampa and unincorporated Hillsborough County zoning.
  • PDs need their own documents. More than a quarter of the vacant parcels carry a PD-type code. A base district table cannot answer what is allowed on them.
  • The model bill is small. Assume about 15,000 input tokens and 1,000 output tokens per parcel. At list price, Claude Sonnet 5.5 costs about $0.04 per check, or about $112 for all 2,808 parcels. Claude Haiku 5.5 costs about $0.002 per check. The build vs buy cost model works this out at 50,000 checks a month.
  • The zoning field is not the zoning map. The Property Appraiser's zoning field is a record, not the official map. Confirm each finalist on the city or county zoning map.

Claude vs ChatGPT for commercial real estate (November 2026)

Both are capable general assistants, and neither knows your parcels. The verified differences that matter for CRE are context size in the apps, long-context API pricing, and fit with your existing tools. This comparison sticks to published product facts and makes no benchmark claims.

ClaudeChatGPT
Context window in the appsUp to 1M tokens, varying by model, on every planReasoning models: 256K tokens on Go and Plus; 400K on Pro
Long PDFs in the appsPDFs up to 1,000 pages; charts and images read on PDFs of 100 pages or fewerAbout 320 pages of text input for reasoning models on Go and Plus; about 680 on Pro
Connect your own dataCustom connectors to remote MCP servers on all plansCustom MCP servers added as plugins, subject to account and workspace policies; developer mode (beta) on Plus and Pro
Reusable workflowsSkills on all plans; ProjectsSkills on all plans; Projects and custom GPTs
Coding agentClaude Code (Pro and up)Codex (limited on Free and Go)
Office toolsExcel, PowerPoint, Word and Outlook on paid plansExcel, Word, PowerPoint and Google Sheets extensions (limited on Free and Go)
Training on your contentOpt-out on individual plans; none by default on Team and EnterpriseOpt-out on individual plans; business plans follow OpenAI's business data terms (check the current terms)
API price, mid-tier modelSonnet 5.5: $2 input, $10 output per million tokensGPT-6.2 Sol: $2 input, $10 output per million tokens
API price, flagship modelFable 5.2: $8 input, $40 outputGPT-6.2 Astra: $8 input, $40 output
API long prompts1M-token window at standard price (Haiku 5.5 charges more above 100,000 tokens)Above 272,000 input tokens, input price doubles and output rises 1.5x
Parcel, zoning or owner dataNone built inNone built in

ChatGPT shows plan prices in local currency at chatgpt.com/pricing, so they are not repeated here. The practical test is simple. Give both the same parcel, code chapter and question, and see which answer cites the right sections. Pick the one your team already works in if the answers are equal.

How Plotzy builds this for your team

Plotzy builds the data side that Claude lacks. These are production systems that connect a model to parcel, zoning and owner data, with every answer citing its code section. Plotzy runs zoning Q&A agents, owner contact research and automated report generation in production today.

What a build includes:

  • Your buy box and criteria, written down as rules a model can apply and a person can audit.
  • Runs across your pipeline or a whole county, not one parcel at a time.
  • Data pipelines from county assessor feeds, your internal systems and third-party APIs. If your team works in Claude, a build can expose that data through an MCP server you control.
  • Outputs into your CRM or spreadsheet.
  • Human review points for PD parcels, split zoning and anything a deal depends on.
  • Citations back to the code section on every answer.

How an engagement runs: a working prototype in the first 2 weeks and a proof of concept within 30 days. Production follows in 8 to 12 weeks, depending on complexity. See custom AI software for the full offer.

What you can do today, self-serve: search 150M+ parcels across 3,200+ US counties in Site Selection, with a zoning category and designation on every parcel. Zoning AI (Production plan and up) answers questions about a parcel's zoning from the ordinance, with the section cited. It covers the Florida jurisdictions where Plotzy has loaded zoning maps and ordinances parcel by parcel. Plotzy's site is also agent-readable. llms.txt lists its pages, and any page returns Markdown when requested with Accept: text/markdown.

If you want Claude working on your parcels with citations your team can check, book a 30-minute call below.

Questions to ask before you connect Claude to deal data

  • Data rights. Does your data license allow parcel and owner records to be sent to a model provider?
  • Freshness. What is the source date of each zoning and ownership field, and when was the code text last amended?
  • Citations. Does every answer point to a section, a record and a date?
  • Review. Who checks PD parcels, overlays and anything a deal depends on?
  • Privacy. Does your plan train on your content by default?
  • Reachability. Can your MCP server be reached on public HTTPS?
  • Cost per run. What does one full screen cost in tokens?

Frequently asked questions

Can Claude look up the zoning for an address?

Not reliably on its own. Claude has no zoning map data and may not be able to read official GIS viewers. Connect it to parcel and zoning data, or upload the record, and ask it to cite the code section.

Is Claude or ChatGPT better for commercial real estate?

It depends on the task and where your team already works. As of November 2026, Claude's apps offer a larger context window (up to 1M tokens) than ChatGPT's (256K to 400K for reasoning models). Run the same parcel question through both and compare the citations.

What are Claude Skills, and how would a CRE team use them?

A Skill is a folder of instructions and optional scripts that Claude loads when a task calls for it. A team might keep its zoning memo format, LOI template and buy box rules as Skills, so every analyst gets the same output.

What is Claude Code used for in real estate?

Claude Code writes and runs code. CRE teams use it to pull county GIS layers, clean parcel exports, join zoning to parcels and build repeatable screens. It is a tool for analysts and engineers, not a data source.

Can Claude read Plotzy's zoning guides?

Yes. Plotzy publishes an llms.txt index, and every page on www.plotzy.ai returns Markdown when an agent requests text/markdown. Parcel-level answers still need a Plotzy account or a custom build.

Sources

Related reading