
Open fact layer for commercial real estate AI agents: deterministic calculators with audit trails
Open fact layer for commercial real estate AI agents: deterministic calculators with audit trails, canonical IDs, a fact schema with provenance, and cited public-record facts. MCP server, REST/OpenAPI, Python SDK. Works with Claude, ChatGPT, Gemini, Copilot, Bedrock. Apache-2.0.
·
An open fact layer for commercial real estate AI agents. Deterministic CRE math with an audit trail, canonical property IDs, and public-record facts that cite the document they came from. Free to use hosted, free to run yourself, open source.
No key, no install:
curl -s https://api.factlot.com/v1/calc/debt_sizing -H 'content-type: application/json' \
-d '{"noi":1000000,"value":15000000,"rate_annual":0.065,"amortization_years":30,"max_ltv":0.65,"min_dscr":1.25}'The answer is a $9,750,000 maximum loan, bound by loan-to-value (the coverage test alone would
allow $10.5 million), with every intermediate step, its formula, and the versioned formula ID
calc.debt_sizing@3 so an agent can cite exactly how the number was made.
Check what your agent read from a lease the same way: it extracts each term with the exact quote
and page, and doc_verify_facts confirms the quote is on that page and states the value, then
returns cited facts or the reason it could not
(document verification). No model runs on FactLot's side and
nothing is stored.
Check us against your own spreadsheet: eight Excel workbooks (office, industrial, retail, hotel, lender, lease accounting and more) put live Excel formulas beside FactLot's numbers with a difference column of 0.00 (synthetic inputs; ARGUS parity is untested and not claimed). What the registry holds, county by county, and the licence and AI-use rights of every source: coverage and rights.
Lease comps, free: 6,994 federal (GSA) lease comps (5.3 MB, public domain) as cited public-record facts, and an open
format with lease_comp_validate so firms can share comps with each other, each with stated rights,
no names, and a derived net effective rent (lease comps). GSA leases are
government-tenant leases, not market comps for every use. Search them from any agent, hosted, with
lease_comp_search.
| Client | How |
|---|---|
| Claude Code plugin | /plugin marketplace add 9mp96ygnbk-alt/factlot, then /plugin install factlot@factlot: twelve CRE skills (loan sizing, lease economics, CAM, ASC 842, returns and more) wired to the hosted tools (plugin) |
| Claude Code | claude mcp add --transport http factlot https://api.factlot.com/mcp |
| Claude (web and Desktop) | Customize, then Connectors, then + Add, then Add custom connector: https://api.factlot.com/mcp. Or open factlot-0.5.0.mcpb |
| ChatGPT | Developer mode app with https://api.factlot.com/mcp, or a Custom GPT Action importing https://api.factlot.com/openapi-3.0.json |
| Cursor, VS Code, other clients | Remote MCP URL https://api.factlot.com/mcp (guides) |
| Any MCP client, run locally | npx -y https://github.com/9mp96ygnbk-alt/factlot/releases/download/v0.5.0/factlot-mcp-standalone-0.5.0.tgz (Node.js 22.6+, no network calls) |
| Python agents | pip install "git+https://github.com/9mp96ygnbk-alt/[email protected]#subdirectory=python" |
| Self-hosted REST and MCP | docker run -p 8787:8787 ghcr.io/9mp96ygnbk-alt/factlot:0.5.0 |
The hosted endpoint is free with no key at 60 requests a minute. The local package, the Desktop extension, the Python install and the Docker image above are release v0.5.0. Their next release brings what the hosted server already runs: the tools registry_lookup_batch and registry_recorded_facts, new inputs on existing tools (registry_search_address address, city, state, postal_code and county_fips; registry_coverage fields; same_building_class, same_local_use_code and local_use_codes on the equity-candidate tools; address and label on the subject and comparables of calc_tx_unequal_appraisal; as_of on the Texas deadline rules), and MCP resources and prompts. report_create, the account tools and the model-backed tools are hosted only. It is listed in the
official MCP Registry
as io.github.9mp96ygnbk-alt/factlot. Runnable examples for REST, Python and agent SDKs are in
examples/.
Try the registry tools on real data: the DC registry sample (23.5 MB, CC BY 4.0) holds 18,077 District of Columbia commercial parcels, 5,513 owning entities, 7,931 cited ownership links, 396 buildings, zoning districts for 16,648 parcels and the latest recorded sale for 12,594.
curl -fLO https://github.com/9mp96ygnbk-alt/factlot/releases/download/registry-sample-dc-2026-11-07/factlot-registry-sample-dc-2026-11-07.db.gz && gunzip factlot-registry-sample-dc-2026-11-07.db.gz
FACTLOT_REGISTRY_DB="$PWD/factlot-registry-sample-dc-2026-11-07.db" npx -y https://github.com/9mp96ygnbk-alt/factlot/releases/download/v0.5.0/factlot-mcp-standalone-0.5.0.tgzPlans: the hosted endpoint is free with no key: every calculator, ID, fact, rules and document tool at 60 requests a minute per IP address, 25 registry lookups a day per IP address, and 3 reports a month per IP address; an optional free key by email counts requests and registry lookups on its own meter (the free keys used from one IP address share 60 requests a minute and 100 lookups a day), and reports stay counted per IP address too. Eight things on the hosted API need a Pro or Team key: value history before the current roll year and superseded facts (registry_get, and the subject of registry_tx_equity_candidates); registry_get with as_of (a record on an earlier date); registry_tx_equity_candidates with tax_year (a roll year you name); a parcel's recorded sales, deeds and mortgages beyond the newest sale or deed (registry_get), and registry_recorded_facts; an owner's parcels beyond the first three (registry_owned_by, and registry_get on an entity); the registry batch tools registry_lookup_batch and registry_tx_equity_candidates_batch; the model-backed tools doc_extract_lease and report_narrative; and the account tools account_status and account_usage. Pro, $49 a month or $490 a year: 1 API key, 600 requests a minute, 5,000 registry lookups a month, 50 reports a month, email support. Team, $599 a month: up to 5 API keys, 3,000 requests a minute, 100,000 registry lookups a month, 1,000 reports a month, priority support, and a vote on which regions open next. A lookup is one call to a registry_* tool that finds something: a record, a resolved ID, owners or search results. Not counted: a call that finds nothing (an empty list, urn null or a NotFound inside loaded coverage; up to an hourly limit), registry_coverage, input errors, misses outside loaded coverage (county, state, entity registry or ID system not loaded) and calls the registry could not answer. Limits are hard caps with no overage charges; at a cap the call returns HTTP 429 with the reset time. Paid plans renew until cancelled (monthly, or yearly on Pro yearly); cancel any time. A registry download licence ($2,400 a year per state file, for internal use, not redistribution, refreshed when the hosted regions are, with no promised interval; see service levels) is available on request, and Enterprise (private deployment, invoicing, a signed data-processing agreement with a no-training commitment, and SSO) is by quote: [email protected]. Terms · Privacy. Questions and bug reports go in GitHub Discussions and Issues.
The hosted API at https://api.factlot.com answers registry lookups on every plan, including the
free tier with no key (25 registry lookups a day per IP address). These regions are hosted:
| Region | Hosted data |
|---|---|
| Maryland | All 24 jurisdictions; 109,249 commercial parcels and 2,812 buildings from SDAT data as of 2026-11-05, public domain |
| The District of Columbia | 18,077 commercial parcels, 5,513 registered business entities and 396 buildings, with owners of record that are organisations, zoning and recorded sales; CC BY 4.0 |
| Juab and Summit counties in Utah | 1,528 parcels, snapshot as of 2025-11-26; CC BY 4.0, "State of Utah, SGID" |
| Eighteen Texas counties | 323,785 commercial parcels: Harris 87,717 with 3,511 buildings, Dallas 59,645 with 998 buildings, Travis 30,556 with 407 buildings, Collin 18,035 with 656 buildings, Cameron 14,173 with 98 buildings, Bell 13,685, Nueces 13,634 with 493 buildings, Denton 13,304 with 311 buildings, Williamson 11,758 with 402 buildings, Galveston 10,088 with 274 buildings, Jefferson 9,594 with 204 buildings, Brazos 8,252 with 64 buildings, Midland 7,641, Ector 7,036, Grayson 6,613 with 41 buildings, Johnson 4,416, Guadalupe 4,081 with 33 buildings and Kaufman 3,557 with 42 buildings; Denton and Williamson from 2027 preliminary rolls as of 2026-11-07; Midland has no values because its publisher's file has none; value history by tax year, organisation owners only, and the 91,641 Texas Comptroller entities that own a hosted parcel; registry_tx_equity_candidates and its batch form answer for these counties; appraisal-district rolls are public record with no licence stated, cited per fact with each source's rights |
| New York City | All five boroughs; 164,633 commercial parcels and 5,071 buildings from the Department of Finance final assessment roll as of 2026-05-15, public record with no restrictions on use, with owners of record that are organisations, and the 66,509 New York Department of State entities that own a hosted parcel |
| Los Angeles County | 203,179 commercial parcels from the Assessor's parcel data roll as of 2026-10-14; no owner names, which the public file does not carry; the County's open-data terms allow commercial use |
| San Francisco | 19,048 commercial parcels from the Assessor's secured roll (the 2025 roll year, the latest published; file as of 2026-06-26); no owner names, which the public file does not carry; public domain (PDDL) |
| The City of Boston | The City of Boston only, not Chelsea, Revere or Winthrop, the rest of Suffolk County; 13,296 commercial parcels from the Assessing Department's fiscal-year 2026 file, with values as of 2025-01-01, older than the other regions; public domain (ODC-PDDL) |
| Denver | 13,539 commercial parcels from the Assessor's file as of 2026-10-29; CC BY 3.0, "City and County of Denver", and the 6,678 Colorado Secretary of State entities that own a hosted parcel |
| Connecticut | 43,398 commercial parcels in eight former counties, from the towns' grand lists in the statewide parcel and CAMA layer as of 2026-11-09; CC0, and the 23,752 Connecticut Secretary of the State entities that own a hosted parcel |
The hosted API serves Maryland (all 24 jurisdictions), the District of Columbia, Juab and Summit counties in Utah, eighteen Texas counties (Harris, Dallas, Travis, Collin, Cameron, Bell, Nueces, Denton, Williamson, Galveston, Jefferson, Brazos, Midland, Ector, Grayson, Johnson, Guadalupe and Kaufman), New York City (all five boroughs), Los Angeles County, San Francisco, the City of Boston, Denver, and Connecticut (statewide, by town): 909,732 parcels in 62 counties. Call registry_coverage first: it lists exactly what is
hosted, and a lookup outside it returns a coverage miss. Every answer cites its source file and the
licence or attribution that goes with it. The registry is built to keep owners that are organisations and to
exclude natural persons' names (if a record names you as a person, or is wrong, email [email protected] and
we will remove or correct it), and the Maryland, Utah, Los Angeles County and San Francisco files carry no owner names. The DC data is the same as the
downloadable sample registry-sample-dc-2026-11-07.
The other regions in the project's pipeline are available only on a self-hosted FactLot with a registry file; more hosted regions open as their terms allow. Tarrant County is not served by the hosted registry while the publisher's terms for AI use are unclear.
Agents doing CRE work get stuck in the same places. They can't trust their own arithmetic, they can't cite where a number came from, and they can't tell that "Suite 400, 123 Main LLC" in one system is the same thing as a parcel number in another.
FactLot gives every agent three things it can use by default:
| Package | What it does | License |
|---|---|---|
@factlot/calc | 26 deterministic CRE calculators: rent schedules, net effective rent, proposal comparison, multifamily rent rolls, percentage rent, CAM reconciliation, capex recovery, loan sizing, amortization, ASC 842 lease remeasurement, prepayment and defeasance, WALT, NPV, IRR, DCF and direct capitalization, equity waterfalls, 1031 exchange dates, lease critical dates, lease vs. own, renew vs. relocate, hotel metrics, Texas unequal-appraisal arithmetic and reconciliation checks. Every result includes an audit trail and a versioned formula ID. | Apache-2.0 |
@factlot/id | FactLot ID gives canonical IDs for parcels, legal entities, buildings, spaces, leases, and documents. IDs are deterministic wherever the world already has a key (parcels, entities, buildings from their street address, spaces, documents), so any agent can compute them offline; leases get minted IDs. | Apache-2.0 |
@factlot/schema | The FactLot data model. A fact is one atomic claim with provenance (source document, page, clause, method) and rights (whether an AI may use or redistribute it). | MIT |
@factlot/tools | The tool catalog, with native tool definitions for Claude, OpenAI, Gemini, Amazon Bedrock, Mistral, and OpenAPI platforms, plus one runTool dispatcher. | MIT |
@factlot/server | The remote connector: Streamable HTTP MCP at /mcp, REST at /v1, and OpenAPI. Runs in Docker. | Apache-2.0 |
@factlot/mcp | An MCP server that exposes all of the above to Claude and any other MCP-capable agent. It runs locally and needs no API key. | Apache-2.0 |
Be clear about scope before adopting it:
https://api.factlot.com serves
the calculators, IDs, fact tools, reports and registry lookups for the regions above (free at 60
requests a minute); registry regions not hosted need a self-hosted copy with a registry file. To run it yourself, install from the
v0.5.0 release, from git, from
source, or with Docker (getting started).| Platform | How to connect |
|---|---|
| Claude | Add a custom connector with https://api.factlot.com/mcp, install the one-click Desktop extension (factlot-0.5.0.mcpb), use Claude Code, or call it from the Claude API |
| ChatGPT and OpenAI | Create a ChatGPT app in developer mode (search/fetch make deep research work), import /openapi-3.0.json (the curated subset of at most 30 operations) as a Custom GPT Action, use the Responses API's remote MCP, or use function calling |
| Google Gemini | Use Gemini API function declarations, Gemini CLI over MCP, or Vertex AI |
| Microsoft Copilot | Add an MCP tool in Copilot Studio, use the Power Platform connector, or build a Microsoft 365 API plugin |
| Amazon Bedrock | Use the Converse API toolConfig, or a Bedrock Agents action group backed by the Lambda handler |
| Mistral, xAI, Llama, DeepSeek, and others | Use OpenAI-compatible function calling or MCP clients |
Every surface serves the same tools, generated from one catalog. See connectors/.
FactLot is not on npm or PyPI yet. These work without either: they run the stdio MCP server from a GitHub release, from a clone, or in Docker. On every release tag the release workflows install each release file the way a user would and test it, and test the Docker image; the last hand-run pass over every install path was made on version 0.1.0 (install verification says what it covered).
# From the GitHub release: one file, dependencies bundled
claude mcp add factlot -- npx -y https://github.com/9mp96ygnbk-alt/factlot/releases/download/v0.5.0/factlot-mcp-standalone-0.5.0.tgz
# From a clone
git clone https://github.com/9mp96ygnbk-alt/factlot.git && cd factlot
npm install && npm run build
claude mcp add factlot -- node "$PWD/packages/mcp/dist/bin.js"
# Remote connector (Streamable HTTP MCP at /mcp, REST at /v1, OpenAPI)
docker run --rm -p 8787:8787 ghcr.io/9mp96ygnbk-alt/factlot:0.5.0Claude Desktop users can install factlot-0.5.0.mcpb from the
release. For any other MCP client, use the
same command in its config:
{
"mcpServers": {
"factlot": {
"command": "node",
"args": ["/absolute/path/to/factlot/packages/mcp/dist/bin.js"]
}
}
}Getting started has every option, including the TypeScript libraries as release tarballs.
The server provides these tools:
calc_rent_schedule, calc_net_effective_rent, calc_compare_proposals,
calc_multifamily_rent_roll, calc_percentage_rent, calc_cam_reconciliation,
calc_cam_reconciliation_batch, calc_capex_recovery, calc_debt_sizing,
calc_lease_remeasurement_asc842, calc_prepayment_and_defeasance, calc_walt, calc_npv,
calc_irr, calc_lease_vs_own, calc_renew_vs_relocate, calc_hotel_metrics,
calc_dcf_pro_forma, calc_direct_cap, calc_amortization_schedule, calc_equity_waterfall,
calc_exchange_1031, calc_critical_dates, calc_reconcile_dcf, calc_reconcile_cam,
calc_tx_unequal_appraisalrules_tx_property_tax_protest_deadline, rules_tx_property_tax_protest_deadline_batchid_parcel, id_entity, id_space, id_building, id_document,
id_parse, fact_validate, fact_package_create, fact_package_verify, lease_comp_validatedoc_verify_facts, doc_abstract_lease, doc_compare, doc_verify_claim,
doc_spread_statement, doc_check_use_conflict, doc_predicatesreport_create builds an XLSX, PDF or ICS file from a tool result, by code with no model, and
returns a download link that works 24 hours (docs/reports.md)search, fetch, schema_getregistry_coverage lists, and on a self-hosted FactLot when
FACTLOT_REGISTRY_DB points at a registry file): registry_get, registry_resolve_parcel,
registry_resolve_external, registry_search_address, registry_search_entities,
registry_owned_by, registry_recorded_facts (every recorded sale and document of a parcel),
registry_coverage, registry_lookup_batch (up to 50 parcels by URN,
account number or address), registry_tx_equity_candidates,
registry_tx_equity_candidates_batchfactlot://coverage,
factlot://sources, factlot://markets, factlot://schemas/tools, factlot://schemas/fact,
factlot://spec/id, factlot://plans; prompts property_owner, owner_portfolio,
parcel_value_history, tx_unequal_appraisal_candidates, lease_abstract_check, deal_math
(docs/agents.md)pip install "git+https://github.com/9mp96ygnbk-alt/[email protected]#subdirectory=python"import factlot
tools = factlot.anthropic_tools() # or factlot.openai_tools(); pass these to your model
# When the model calls a tool, run it by name with the model's arguments:
result = factlot.run_tool("calc_npv", {"cash_flows": [-1000, 600, 600], "rate": 0.08})
result["value"]["npv"] # 69.96The Python SDK has no dependencies and passes the same conformance vectors as the TypeScript packages. A differential test runs random inputs through both SDKs in CI and fails on any difference, down to the last digit. See python/README.md.
import { debtSizing } from "@factlot/calc";
import { parcelId } from "@factlot/id";
const loan = debtSizing({
noi: 1_000_000,
value: 15_000_000,
rate_annual: 0.06,
amortization_years: 30,
max_ltv: 0.65,
min_dscr: 1.25,
min_debt_yield: 0.09,
});
loan.value.max_loan; // 9750000
loan.value.binding_constraint; // "ltv"
loan.steps; // every intermediate value, with formulas
loan.formula; // { id: "calc.debt_sizing", version: 3 }
parcelId({ stateFips: "06", countyFips: "037", apn: "5432-001-012" });
// { kind: "parcel", urn: "fl:parcel:us-06037:5432001012", short: "flp_..." }This requires Node.js 22.6 or newer.
npm install
npm run check # lint, format check, typecheck, tests, buildFactLot is open source and fully self-hostable, which matters when leases are confidential.
@factlot/schema, @factlot/tools and the client SDKs are licensed under MIT."FactLot" and the FactLot logo are trademarks of their owner. The open-source licenses cover the code, not the name.
Pick your client and paste the snippet. Each one is the same server, written the way that client expects it.
claude mcp add factlot -- node /absolute/path/to/factlot/packages/mcp/dist/bin.js{
"mcpServers": {
"factlot": {
"args": [
"/absolute/path/to/factlot/packages/mcp/dist/bin.js"
],
"command": "node"
}
}
}code --add-mcp '{"name":"factlot","args":["/absolute/path/to/factlot/packages/mcp/dist/bin.js"],"command":"node"}'[mcp_servers.factlot]
command = "node"
args = ["/absolute/path/to/factlot/packages/mcp/dist/bin.js"]Runs locally on your device. Your client starts the server itself, so nothing has to be hosted.
Paste this prompt into your agent. It reads this page and does the setup for you.
Read https://aiagentslisting.com/mcp/factlot to learn what the "FactLot" MCP server does and how to install it. Add it to my coding agent's MCP configuration as documented on that page, then confirm the server connects and list the tools it exposes.Agents can also browse this directory over MCP at https://aiagentslisting.com/api/mcp
claude mcp add factlot -- npx -y https://github.com/9mp96ygnbk-alt/factlot/releases/download/v0.5.0/factlot-mcp-standalone-0.5.0.tgz
claude mcp add factlot -- node "$PWD/packages/mcp/dist/bin.js"docker run --rm -p 8787:8787 ghcr.io/9mp96ygnbk-alt/factlot:0.5.0pip install "git+https://github.com/9mp96ygnbk-alt/[email protected]#subdirectory=python"This server runs locally, so we can't read its tool list over the web yet.
We could not run this server: its endpoint did not list its tools to an anonymous client, which usually means it asks for an API key or a sign-in. The four checks that need a live server are left out, and the grade tops out at B until we can read its tools.
Maintain this server? Claiming proves you own the listing and earns the last 4 points.
Claim this listingPaste this line near the top of your repository's README. The badge always renders the current score.
[](https://aiagentslisting.com/mcp/factlot)Nothing comparable is listed yet.
v0.5.0FactLot 0.5.0Release notesregistry-sample-dc-2026-11-07Registry sample: District of Columbia (2026-11-07)Release notesv0.4.0FactLot 0.4.0Release notesv0.3.0FactLot 0.3.0Release notesv0.2.0FactLot 0.2.0Release notesregistry-sample-dc-2026-11-04Registry sample: District of Columbia (2026-11-04)Release noteslease-comps-gsa-2026-11-04Lease comps: GSA federal leases (2026-11-04)Release notesv0.1.1FactLot 0.1.1Release notesv0.1.0FactLot 0.1.0Release notesregistry-sample-dc-2026-11-03Registry sample: District of Columbia (2026-11-03)Release notesConnect to the AI Agents Listing MCP server at https://aiagentslisting.com/api/mcp using the HTTP transport and complete the sign-in when it asks. Then call the submit_review tool for the listing "factlot" with a rating from 1 to 5 and a short note on what worked and what did not.No reviews yet
Be the first to share your experience with this tool.
One email a week. New agents, MCP servers and skills, and what is actually getting traction.