add_deal_note
ChatGPTAdd an idempotent analyst-visible note to an accessible EQUIRE deal. Requires deal write access; does not alter review/release state.
add_listing_to_pipeline
ChatGPTGOVERNED: saves a global CRE listing to a SPECIFIED prospecting mandate's Screening Queue (a reviewable prospect). It does NOT create a deal; conversion to a deal happens in EQUIRE through the existing convert gate. Requires an org write role (viewers denied). The write is idempotent by the listing's source URL (idempotency_key is recorded for intent/audit, not the dedup key).
attach_deal_document
ChatGPTAttach one original supported document from a downloadable file reference to an existing deal and queue canonical processing. Requires explicit confirmation and documents:write. Content-hash duplicate detection is non-atomic; processing and source verification are reported separately from file acceptance.
attach_deal_photo
ChatGPTAttach one PNG, JPEG or WebP property photo from a downloadable file reference to the deal gallery. Requires explicit confirmation and documents:write. Camera EXIF/GPS metadata is stripped. This operation is not idempotent; an uncertain response does not establish whether a photo was saved.
attach_research_sources
ChatGPTFetch 1-5 report URLs and link them to 1-10 accessible deals' research. Attaching links the source; its figures are read by the next research run (run_market_research) or Continue fact review, so a linked_only outcome is expected here and is not a failure. Web pages need hosted reading, which may be turned off on a deployment; PDF links are supported, though a PDF can still be unreachable. A source already linked to the deal reports already_linked without a refetch, but the batch is non-atomic: after an uncertain response, pairs not reported as linked may or may not be linked. All targets and confirmed scope versions are validated before fetching. Only URLs are accepted; assistant-authored report prose is not. Valuation values remain proposals.
compare_deals
ChatGPTCompare 2-5 accessible deals side by side across economics (price, $/SF, cap rate; each row's pricing carries the release basis of its price and cap rate), diligence and health counts, and — with valuation:read — headline valuation returns. Inaccessible deal ids are returned as accessible:false rather than failing the call.
complete_deal_file_upload
ChatGPTComplete one private upload after its raw PUT succeeds. Rechecks current deal permission and billing, verifies original size and SHA256, consumes the bound ticket once and attaches through canonical document processing or photo sanitization. A consumed ticket is not a replay receipt; after an uncertain outcome inspect Documents or the gallery before starting a new upload. Source review and release remain separate.
create_deal
ChatGPTCreate a standalone deal workspace in one connected organization with assistant-supplied identity. Requires explicit confirmation, an organization write role and deals:write. The same idempotency_key resolves the same deal. Financial facts and source verification are not created by this operation.
create_diligence_item
ChatGPTCreate an idempotent due-diligence checklist item on an accessible deal. Completion, waiver, or flagged status requires a reason or notes.
fetch
ChatGPTFetch a compact, permission-checked deal summary, document excerpt, CRE listing, or retained organization website source by the id (or equivalent equire:// uri) returned on a search result. Deal/document fetches need deals:read/documents:search; listing fetches need listings:read; organization website sources and generations need organization_website:read. Document source_independence is derived, independent, or unavailable. The legacy derived_from_supporting_analysis field is a non-independence alias for any stored derivation, including other EQUIRE outputs; it does not attest their origin. Derived or unavailable sources cannot establish independent corroboration.
get_capital_stack
ChatGPTCapital stack for an accessible deal: debt and equity layers with their terms, covenant coverage (null when the engine could not measure it or the model is all-cash, never a sentinel) and the financing basis. A deal with no saved model or analysis year reads no_valuation_saved or missing_analysis_basis, never empty inputs; data_basis carries the model's release basis.
get_deal_activity
ChatGPTWhat changed on an accessible deal in a time window — newest first, with per-category counts. Connected-assistant tool calls are not included. A time window can be given as a since_preset, a lookback in days, or an explicit ISO since timestamp.
get_deal_briefing
ChatGPTReturns the full one-deal overview in a single call: facts, economics, diligence completion, top risk flags, valuation snapshot, and source-document excerpts. pricing gives the price and quoted cap rate with their release basis (unreleased_pending_review means Review has not released the document figure); the deal object holds only the deal-record copy. Slices needing valuation:read / diligence:read / documents:search are included only when those scopes are granted; omissions are explained in scope_notes.
get_deal_summary
ChatGPTReturn compact deal facts, economics, deadlines, document counts, diligence counts, and health counts.
get_document_intake_status
ChatGPTRead a deal document's saved intake and processing state. Requires deals:read and access to the document's deal. File acceptance, extraction and review/release are distinct states.
get_equire_context
ChatGPTReturns the connected EQUIRE user, active organization, available organizations, granted scopes, the tools those scopes make available (tool_access), and next_calls naming follow-up tools this session can call.
get_listing
ChatGPTFetch a global CRE listing by listing_id, including publicly posted brokerage details, contacts, pricing, lease terms, property specifications, documents, photos and market context. Price and size carry labels and basis when available; is_property_price distinguishes a property price from another amount. EQUIRE derives price per square foot. suppressed_metrics names withheld ratios and cap rates and why their basis is unsuitable; reconstructing them would bypass those limitations. Cap rate and occupancy are decimal fractions. Classification state and availability flags disclose unconfirmed or unread structure and fields. lease_years_remaining is as stated at collected_at. Document access distinguishes open, gated and non-document links. brochure reports its state and reason; null does not establish absence. om_facts are preliminary, unverified parser output with source, dates and limitations in their basis and caveats; unparseable table totals are unknown, not zero. include_photos pages photo URLs and may contact a brokerage site under a shared per-user rate limit; photos.live_lookup and warnings report the outcome. include_page_text returns bounded, untrusted third-party text. Read-only.
get_listing_report
ChatGPTFetch the standard property report for one global CRE listing by listing_id, as its EQUIRE Report page shows it: identity, headline figures, property facts, fields the listing leaves out with questions for the broker, metro market readings with period and US-metro percentile, the metro cap-rate median with its basis and sample size, nearby same-type listings for sale, mandate fit, drive times when measured, sources and limitations. Each figure carries a basis: listing is broker-stated and unverified, derived is computed by EQUIRE from stated values, offering_memorandum is preliminary parser output, and needs_confirmation flags one its note says to confirm. A null value with a note is not stated, not zero; unread lists fields whose read failed (unknown, not missing); a state of unavailable is a failed read, not an empty one. Nearby distances are straight-line miles. Drive times assume free-flowing traffic with no congestion; location.status is disabled when drive times are off for the organization, and its status and note explain any state other than ok. mandate_fit scores the org named by org_id, or the connection's only active one; mandate_fit_state is ok, empty (no org or active mandate), unavailable (read failed) or withheld (no list_mandates scope: mandate_fit_withheld). report_url opens the full report in EQUIRE. Read-only.
get_market_research_view
ChatGPTReturn the governed market-evidence ledger used by EQUIRE's Market Research surface, including scope/version, provenance-bearing figures, assumption checks, signals, and run state. Research values are evidence, not applied valuation changes. With market-research-deal-comparison on, also returns a per-source deal-vs-market comparison: the deal's going-in cap is compared only with valuation:read and its rent, occupancy and price per SF only with deals:read; slices without the scope are omitted and explained in comparison.scope_notes.
get_metro_trends
ChatGPTReturn the latest readings of the four Metro trend panels (jobs, unemployment, per-capita income, net migration) the Market Research Metro tab shows: each panel's latest value and period, how its benchmark reads at the period both lines share, and the ranked-peer median. Latest readings only, not the series. A trend is context and carries no deal verdict. The benchmark is the Median metro or the US payroll series, as each statement names it.
get_operating_expenses
ChatGPTOperating expense inputs for an accessible deal: the saved OpEx lines and scalars with their source, the lines the selected scenario's DCF prices, and year-1 recoveries. A deal with no saved model or analysis year reads no_valuation_saved or missing_analysis_basis, never empty inputs; data_basis carries the model's release basis.
get_organization_agent_findings
ChatGPTBrowse compact saved finding metadata, including investment facts and not-verified findings. Read one findingKey and named section/offset for complete evidence. Preserve own/outside-site and proof labels; this never starts research or spends.
get_organization_investment_history
ChatGPTRead bounded public, source-reported investment claims, counts and conflicts for one connected organization. These claims are not verified deal facts or ownership records.
get_organization_research_notes
ChatGPTRead the web research notes recorded for one connected organization, newest first: each note's question, brief answer, the source page's exact words and address, whether the page is the organization's own site or an outside site, and whether code confirmed the quote on that page. Outside-site notes are that site's claims; unconfirmed notes are not verified.
get_organization_source_history
ChatGPTRead a retained source's bounded revision-fenced generation and observation history. Source bodies are not returned by this history tool.
get_organization_summary
ChatGPTReturn a permission-checked organization summary with public/company profile, compact firm memory, portfolio counts, recent deals, and recommended drill-down calls.
get_portfolio_rollup
ChatGPTPortfolio-wide health roll-up across accessible deals (optionally one org): open_total and open_by_severity count OPEN issues, all_by_status is the status distribution across ALL issues, plus deals ranked by risk and — with diligence:read — aggregate diligence completion. With valuation:read and one active-membership org_id, it also returns governed value, returns, and risk metrics with per-metric basis disclosure.
get_rent_roll
ChatGPTTenant model inputs for an accessible deal, paginated: name, suite, leased SF, in-place rent, lease dates and WALT inputs, with each rent-roll fact's source under the bound release and analyst overrides reported separately from source values. A deal with no saved model or analysis year reads no_valuation_saved or missing_analysis_basis, never empty inputs; data_basis carries the model's release basis.
get_research_run
ChatGPTRead the latest durable market-research run for a deal, or one exact run_id. Returns progress, bounded counts, and sanitized failure state.
get_review_item
ChatGPTRead one open Review item on a deal: its candidate values with their source locators (document, page or cell), the analyst-entered value as separate authority, grouped signals, and release basis. Verbatim document quotes are included only with documents:search; scope_notes explains an omission.
get_valuation_snapshot
ChatGPTReturn the latest saved valuation assumptions and returns for an accessible EQUIRE deal.
inspect_organization_website_catalog
ChatGPTInspect bounded catalogued website URLs, folders and pending admission states. Discovery only; this does not fetch or retain pages.
list_deal_health_issues
ChatGPTList current health issues, red flags, stale items, and blockers for a deal or accessible portfolio.
list_deal_notes
ChatGPTList MCP-origin notes for an accessible deal, without exposing unrelated deal communications.
list_deals
ChatGPTList accessible deals with structured filters, sort controls, and pagination for portfolio browsing. Filters apply to deal records, not document text.
list_diligence_items
ChatGPTList due diligence checklist items for a deal with optional status, category, and priority filters.
list_mandates
ChatGPTList prospecting mandates in the connected organizations: mandate_id, name, organization, status, and saved result count (null with result_count_limitation when the count could not be read). mandate_id is the id add_listing_to_pipeline takes and the search_profile_id on list_prospects rows.
list_organization_updates
ChatGPTBrowse compact retained own-site news and update metadata. Publication and retrieval are distinct dates; filters apply before paging. Read one article with fetch using uri equire://organization-source/<org_id>/<sourceId>, then body_offset/expected_revision for complete continuation.
list_prospects
ChatGPTList accessible prospecting results and their result_id values so prospect_sources_for can inspect evidence sources.
list_review_items
ChatGPTList one deal's EQUIRE Review items in Review priority order, with the Review page's counts and release-gate approval state. Each row's review_item_id is the id stage_review_resolution takes. status open (default) is the Review page's queue, blocking is the decisions that clear a release blocker, all adds supporting findings.
list_staged_review_resolutions
ChatGPTList MCP-origin staged review-resolution proposals for an accessible deal. These are pending proposals only, not applied review changes.
list_staged_valuation_updates
ChatGPTList MCP-origin staged valuation assumption updates for an accessible deal. These are pending proposals only, not applied valuation changes.
open_equire_workspace
ChatGPTOpen the governed EQUIRE deal workspace from the sidebar or conversation. Select an organization, browse accessible deals, inspect scope-aware briefings, and retrieve source evidence. Optional deal_id opens an accessible deal; org_id selects one authorized organization. This tool performs no workspace writes.
prepare_deal_file_upload
ChatGPTPrepare a private object-specific upload for an original document or property photo after confirmation. Returns a raw PUT URL, required headers and a 15-minute completion ticket. A sandbox can PUT bytes without publishing a file or forwarding EQUIRE credentials; sandbox egress approval may be required. The signed upload URL itself remains valid for 2 hours. Storage is required; there is no process-memory upload fallback.
preview_debt_sizing
ChatGPTRead-only preview of the largest senior loan the deal supports under lender constraints (max LTV, max LTC, min DSCR, min debt yield). Returns the supportable proceeds, which constraint binds, each constraint's cushion, the implied equity and a sources and uses tie-out, sized on the saved Base scenario's NOI. Provide at least one constraint. Ratios are decimal fractions (65% LTV = 0.65); DSCR is a multiple (1.25). NOTHING IS SAVED OR APPLIED: applying a sized loan happens in EQUIRE, where an analyst approves it. Unmeasurable ratios (for example coverage on zero proceeds) are null.
preview_market_research
ChatGPTResolve a bounded brokerage-report research plan for 1-10 accessible deals from each deal's confirmed market scope. This performs no web search or fetch. Returns a server-signed plan per deal; nothing runs until run_market_research receives it.
prospect_sources_for
ChatGPTReturns all evidence sources (URLs, PDFs) attached to a specific prospect search result. Accepts resultId or result_id from list_prospects.
recover_market_report_sources
ChatGPTRetry 1-10 explicitly named market-report cache IDs in one deal's exact current research scope. Requires an idempotency key, expected scope version, and audit reason. Reuses the existing durable research Workflow and receipt owner; never resets attempts, broadens sources, searches, or changes scope.
research_organization_question
ChatGPTResearch one question about a connected organization on the public web: an agent searches, reads the organization's own pages and outside pages, and records each answer with the source page's exact words. Organization administrators only. It spends the organization's AI allowance (many searches and page reads) and saves the notes to the organization's record, so confirm with the user first, stating the question. Each note names its page, is labelled own site or outside site, and says whether code confirmed its quote.
run_market_research
ChatGPTStart durable brokerage-report research for 1-10 deals from server-signed plans returned by preview_market_research. This spends search/PDF budget and links evidence, but never applies valuation changes; its results are research evidence, not approved underwriting assumptions.
run_sensitivity
ChatGPTRead-only what-if. Sweeps each named assumption one at a time across 5 points centred on the deal's current value, through the deal's own DCF, and returns the levered IRR and equity multiple at each point plus a ranking by IRR spread (widest first), which answers 'which assumptions matter most'. NOTHING IS SAVED. Up to 8 variables per call. Rate-shaped values are decimal fractions (6.5% = 0.065). discountRate is unsupported (it moves value, not IRR), and interestRate is unsupported when the senior debt floats; unsupported variables are reported, not dropped. A null IRR or multiple means that point has no meaningful solution.
search
ChatGPTSearch accessible EQUIRE deals and optional document excerpts. Document source_independence is derived, independent, or unavailable. The legacy derived_from_supporting_analysis field is a non-independence alias for any stored derivation, including other EQUIRE outputs; it does not attest their origin. Derived or unavailable sources cannot establish independent corroboration.
search_deal_documents
ChatGPTSearch processed EQUIRE document chunks for a deal and return bounded excerpts with source metadata. Document source_independence is derived, independent, or unavailable. The legacy derived_from_supporting_analysis field is a non-independence alias for any stored derivation, including other EQUIRE outputs; it does not attest their origin. Derived or unavailable sources cannot establish independent corroboration.
search_documents
ChatGPTSearch processed document excerpts across all accessible deals and return bounded results with source metadata. Requires documents:search scope. Never returns full documents; only excerpt text with deal, document, page, and section provenance. Document source_independence is derived, independent, or unavailable. The legacy derived_from_supporting_analysis field is a non-independence alias for any stored derivation, including other EQUIRE outputs; it does not attest their origin. Derived or unavailable sources cannot establish independent corroboration.
search_listings
ChatGPTSearch EQUIRE's GLOBAL indexed CRE listing inventory (sale and lease listings collected from publicly posted brokerage listings), NOT your own deals. Listing details and broker contacts come from listings brokerages publicly post. Returns structured rows (price, $/SF, cap rate as a decimal, lease rate, size, location, source brokerage, brokerage source_url, primary_broker with name, title, email, and phone plus broker_count, document/image availability, om_filled for the fields filled from the offering memorandum, om_withheld for the fields left empty because that memorandum could not be matched to the listing) plus a citable url and a listing_id for get_listing / add_listing_to_pipeline. has_documents is true only when the listing has a document worth opening, never counting a buyer portal, deal room, agreement, floor plan or site boilerplate link, and is null when the links could not be read. $/SF is EQUIRE-derived (price over stated size, land over land area) and withheld outside a plausible range; a lease row carries no sale price; suppressed_metrics names each withheld figure and why, and structure_suppressions counts them for the page. om_filled and om_withheld are null when the offering memorandum facts could not be read. A null field named in om_withheld is not a report that the memorandum is silent: get_listing shows that figure with a caveat. Each row also carries headline_amount (the amount with its label and price_basis), interest_kind, asset_scope, price_basis, size_basis, structure_tags, classification_state (unassessed, needs_review or classified; unread when the classification lookup failed, not_applicable for a lease, disabled when classification is switched off), the gates psf_allowed, per_unit_allowed, cap_rate_allowed and market_evidence_eligible, and listing_contract_version (per row); structure_note says what a state without a confirmed classification means and makes no structure claim, and needs_review rows keep their labels while a person has not yet confirmed them. sale_price_usd keeps its value beside is_property_price, which is read from the amount shown (an OM-filled asking price on an unclassified row is a property price): when is_property_price is false the amount is a loan price, unpaid balance, starting bid, raise amount or other non-property amount named by price_label, and it is not an asking price. Quote price_label with the amount and size_label with the size, and never derive a ratio EQUIRE withheld. While classification is on (classification_state is not disabled), min_price and max_price match property prices only, so a loan price, starting bid or raise amount is never inside a price range, and a price sort or cap_desc places such a listing after every ranked row without ranking it; with classification off both use the stated amounts. excluded_by_price_basis counts what that did across the whole result set, not the returned page: price_filter_excluded listings were removed by the price filter, while price_sort_listed_last and cap_rate_sort_listed_last listings stay in the result set after every ranked listing, unranked; results is one board page (at most three listings per source), so they are normally not in it. The object is absent when no count was taken (including with classification off), which means not counted, never zero. Read-only.
search_mentions
ChatGPTSearch authorized deals for the composer. Returns at most 20 suggestions across at most 20 consented organizations.
search_organization_website
ChatGPTSearch cited retained public website passages for the current organization. Uses the organization's AI allowance.
simulate_valuation_scenario
ChatGPTRead-only what-if. Evaluates hypothetical assumption changes through the deal's own DCF model and returns baseline vs simulated returns plus the change in each. NOTHING IS SAVED — the deal's assumptions, scenarios, and saved valuation are untouched; the result is not an applied change (stage_valuation_assumption_update records a staged proposal for human approval). Each adjustment takes EITHER a multiplier (1.05 = +5%) OR a target_value (the absolute number to move to). Rate-shaped assumptions are decimal fractions: a 6.5% exit cap is 0.065, not 6.5. Each adjustment reports a status and baseline_basis: bounded targets are not fulfilled, and input_only means the input moved without an isolated economic effect. Unmodellable keys are reported rather than silently dropped.
stage_review_resolution
ChatGPTStage a proposed review-item resolution for EQUIRE approval. This does not resolve conflicts, release data, or make IC decisions.
stage_valuation_assumption_update
ChatGPTStage a valuation assumption change for EQUIRE review. This does not apply the model change or alter released deal data.
update_diligence_item
ChatGPTUpdate selected fields on an existing due-diligence item. Completion, waiver, or flagged status requires a reason or notes.