account_bundle_delete
ChatGPTRemove one saved bundle so new quotes stop getting it. Inspections that already have it keep it, priced exactly as quoted, and the result says how many. A direct request naming the bundle authorizes it.
account_bundle_list
ChatGPTRead the inspector's saved service bundles: each one's services, its price (amount off or price together) and a plain summary such as "Home Inspection + Radon Testing: $50.00 off together ($500.00, $550.00 separately)". Call this before changing a bundle so you use its exact name. A bundle applies once per inspection, only when ALL its services are on it; when bundles overlap the biggest saving wins; a manual price on a service takes it out of its bundle.
account_catalog_list
ChatGPTRead the inspector's saved services, fees, and discounts — the menu they pick from when building a job, not what is on any one job. Call this before saving anything so you do not create a second copy of something already there, and before applying a fee or discount to an inspection so you use the exact saved name. Fees and discounts come back as percent or amount, never a bare number.
account_discount_create
ChatGPTSave one discount to the inspector's catalog so it is available on future jobs — a repeat-client rate, a military discount. Give percent OR amount, never a bare number. Call account_catalog_list first: an active discount with this name is UPDATED rather than duplicated. To charge one specific inspection, use inspection_fee_apply or inspection_discount_apply instead. This changes the catalog only. Inspections that already have it keep the exact price they were quoted, and no client is emailed.
account_discount_delete
ChatGPTTake one discount off the inspector's menu so it stops appearing on new jobs. Inspections that already carry it keep it, exactly as quoted, and the result says how many. A direct request naming the catalog discount authorizes it; when the removal is your idea, say it affects future jobs only, and a yes executes it. Use a fresh account_catalog_list result; no separate confirmation follows.
account_discount_update
ChatGPTChange one saved discount: rename it or change what it takes off. Give percent OR amount to change the value; omit both to leave it alone. To charge one specific inspection, use inspection_fee_apply or inspection_discount_apply instead. This changes the catalog only. Inspections that already have it keep the exact price they were quoted, and no client is emailed.
account_fee_create
ChatGPTSave one fee to the inspector's catalog so it is available on future jobs — a service charge, a state tax, a travel fee. Give percent OR amount, never a bare number. Call account_catalog_list first: an active fee with this name is UPDATED rather than duplicated. To charge one specific inspection, use inspection_fee_apply or inspection_discount_apply instead. This changes the catalog only. Inspections that already have it keep the exact price they were quoted, and no client is emailed.
account_fee_delete
ChatGPTTake one fee off the inspector's menu so it stops appearing on new jobs. Inspections that already carry it keep it, charged exactly as quoted, and the result says how many. A direct request naming the catalog fee authorizes it; when the removal is your idea, say it affects future jobs only, and a yes executes it. Use a fresh account_catalog_list result; no separate confirmation follows.
account_fee_update
ChatGPTChange one saved fee: rename it, change what it charges, or change what it is charged on. Give percent OR amount to change the value; omit both to leave it alone. To charge one specific inspection, use inspection_fee_apply or inspection_discount_apply instead. This changes the catalog only. Inspections that already have it keep the exact price they were quoted, and no client is emailed.
account_pricing_modifiers_list
ChatGPTRead the inspector's automatic pricing modifiers from Settings > Service & Pricing: square-footage bands, year-built tiers, foundation surcharges and mileage bands, each tier as a range label, its upTo handle and the dollars it adds. Also says whether automatic pricing is on, which saved services apply modifiers (with their base prices, which every tier is added to) and which do not (the tiers add nothing to those until modifiers are turned on for them), nextSteps to offer, and whether distance pricing has a home base. Read this before changing any tier.
account_service_create
ChatGPTSave one service to the inspector's catalog so it is available on future jobs. Call account_catalog_list first: if an active service already has this name, this UPDATES it rather than creating a second one, and you must tell the inspector it was an update. To put a service on one existing job, use inspection_service_add instead. This changes the catalog only. Inspections that already have it keep the exact price they were quoted, and no client is emailed.
account_service_delete
ChatGPTTake one service off the inspector's menu so it stops appearing on new jobs. Inspections that already have it KEEP it, priced exactly as quoted, and the result says how many. A direct request naming the catalog service authorizes it; when the removal is your idea, say it affects future jobs only, and a yes executes it. Use a fresh account_catalog_list result; no separate confirmation follows.
account_service_update
ChatGPTChange one saved service: rename it, reprice it, change how long it takes, or change whether it is charged per property or per unit. Address it by the name the inspector uses. This changes the catalog only. Inspections that already have it keep the exact price they were quoted, and no client is emailed.
automation_create
ChatGPTAdd an automation: either a preset by its preset key from automations_list (this also brings back a removed built-in), or a new one from a full definition and a name. It starts switched on unless active is false. Errors are never saved: a refusal lists what to fix. Automations send to real clients and agents later, on their own, so this pauses for one confirmation first.
automation_explain
ChatGPTExplain one automation in plain words: when it sends, the conditions, each email or text step, and who it would reach in the next 14 days with the reason anyone is skipped. Checked against real inspections; sends nothing. Use it to answer "what does my review request do" or before suggesting a change.
automation_send_log
ChatGPTRead the automation send log: each automated email or text that went out, is scheduled, or was skipped, with the reason for each skip (for example "Agent was asked Aug 14", "No mobile number on file", "Automation paused"). Filter by automation and/or inspection. Use it to answer "why did or didn't my client get the reminder".
automation_set_status
ChatGPTSwitch an automation on ("active") or pause it ("paused"). Pausing stops it at once, including anything it already scheduled. Pausing a built-in means clients and agents stop getting that message. Automations send to real clients and agents later, on their own, so this pauses for one confirmation first.
automation_test_send
ChatGPTSend a test of one automation, or with all: true of every switched-on automation ("test all"), to the inspector only: their account email, and their phone for text steps. Each test is marked [Test] and filled in with their latest inspection. Tests arrive within about 5 minutes; test all can run once every few minutes. No client or agent ever gets a test. It still sends real messages, so it pauses for one confirmation first.
automation_update
ChatGPTChange an automation: its name and/or its whole definition (timing, conditions, steps). Read it first and send back the complete edited definition. Changes apply to messages it sends from now on, including ones already scheduled. Errors are never saved; the last saved version keeps sending. Automations send to real clients and agents later, on their own, so this pauses for one confirmation first.
automations_list
ChatGPTList the inspector's email and text automations (Settings > Automations): each one's name, stage, whether it is on or paused, and a one-line summary of when, if and what it sends, plus presets they can add. Built-in automations are today's messages (invitations, reports ready, review requests...). Until one is edited it sends exactly as before. Use this before changing, explaining, pausing or testing one.
contact_create
ChatGPTSave one person to the inspector's contact book. Call contacts_search first: if someone with the same email and type is already saved, this UPDATES them rather than creating a second entry, and you must tell the inspector it was an update. Saving a contact sends them nothing. This is the inspector's saved contact book, not the people on a job. To put someone on an inspection, use inspection_clients_update.
contact_delete
ChatGPTRemove one person from the inspector's contact book. Inspections they are already on KEEP them, with their name and email intact — only the link back to the book is lost, and the result says how many inspections that affected. A direct request naming the saved contact authorizes it; when the removal is your idea, say it only unlinks the book entry, and a yes executes it. Use a fresh contacts_search result; no separate confirmation follows. This is the inspector's saved contact book, not the people on a job. To put someone on an inspection, use inspection_clients_update.
contact_update
ChatGPTChange one saved contact: rename them, or set their company, phone, email, or notes. Address them by the name the inspector uses, and pass type when two saved contacts share a name. Send null for a field to clear it. Changing a contact sends them nothing. This is the inspector's saved contact book, not the people on a job. To put someone on an inspection, use inspection_clients_update.
contacts_search
ChatGPTSearch the inspector's saved contact book by name, email, or company, optionally narrowed to one type. Read this before saving a contact so you do not create a second copy of someone already in the book, and before changing or removing one so you use the name they actually stored. This is the inspector's saved contact book, not the people on a job. To put someone on an inspection, use inspection_clients_update.
email_message_get
ChatGPTRead one built-in email's subject line and body (HTML, with {{fields}} such as {{client_name}} and {{address}}) before suggesting or making a change.
email_message_update
ChatGPTChange what a built-in email says: its subject line and/or its bodyHtml, or put it back to the app's wording with reset: true. Read it with email_message_get first; leave out what stays the same. Only fields valid for that email are allowed ({{client_name}}, {{address}}, {{inspection_date}}, {{inspector_name}}, {{company_name}}...); a refusal says which one is wrong and nothing is saved. Every email sent from now on uses the new wording, so this pauses for one confirmation first.
email_messages_list
ChatGPTList the built-in emails clients and agents get (Settings > Emails and texts): each one's email key, name, who it goes to, when it is sent, its subject line, and whether the inspector changed the wording. Every inspector has these, with or without automations. Use email_message_get to read a body.
final_checks_get_status
ChatGPTRead a bounded history of Final Checks runs for the current inspection, plus the latest run, progress, failures, and staleness. Each history run has a reviewState and issue dispositions by category: outstanding means PENDING inspector review; resolved is accepted/reviewed plus dismissed; accepted and dismissed remain separate. OUTSTANDING is actionable on the current report, RESOLVED means every finding in that run has an inspector decision, CLEAR means it found nothing, and SUPERSEDED or STALE findings are historical rather than a current review queue. Use this before an ordinary Final Checks start and whenever the inspector asks what has already run, whether it is still running, or what remains to review. Use the returned persisted status rather than earlier chat messages: READY_FOR_REVIEW has results ready, COMPLETED is finished, and FAILED or CANCELLED is not running.
help_search
ChatGPTSearch Hometrace help docs and published release notes. Call for feature, settings and how-to questions before claiming Hometrace cannot do something. Answer from matches and link the doc.
inspection_bundle_remove
ChatGPTTake one service bundle off this inspection, so its services go back to their own prices. The removal sticks: the bundle does not come back on its own. Read inspection_services_list first for the bundle's name and saving, and tell the inspector the price goes up by that saving. Nobody is emailed. A sent or paid invoice cannot change; the tool says so.
inspection_bundle_restore
ChatGPTPut back a service bundle the inspector removed from this inspection. It applies again only if all its services are still on the job and no bigger bundle covers them; the result says whether it applied. Nobody is emailed. A sent or paid invoice cannot change; the tool says so.
inspection_clients_update
ChatGPTAdd, update, or remove the clients or real-estate agents attached to an existing inspection. Read the current contacts from inspection_get_context first and use their contactId values. Adding or changing a contact with an email immediately sends that person an invitation by email, and by text when they have a phone number. Call it for a clear request; when anyone would be invited, nothing is saved or sent until the inspector taps Send on the card it returns. On a draft it invites nobody: invitations go out when the inspector creates the inspection, so never offer to send them.
inspection_details_update
ChatGPTChange the property address or details of an existing inspection: address, city, state, zip, home type, foundation, year built, square footage, purpose, or notes. Changing the address or purpose can make an already-signed agreement stale, and changing square footage, foundation, or year built can change the quoted price. Execute a clear user-authored request only when the exact replacement values and any applicable agreement or quote effect are named; otherwise ask one concrete question for the missing effect, then execute the completed request without a separate confirmation. This call sends no email or text.
inspection_discount_apply
ChatGPTPut one of the inspector's saved discounts on this inspection, by its name. Read inspection_services_list for the available discounts. This changes the quoted price and emails every client attached to the job that it changed. When anyone would be emailed, nothing changes until the inspector taps Send on the card this returns; voice cannot send it. On a draft it emails nobody.
inspection_discount_remove
ChatGPTTake one discount off this inspection, raising what the client owes. The discount stays in the inspector's saved list and can be applied again; only this job loses it. Refused when the discount is not on the job. This changes the quoted price and emails every client attached to the job that it changed. When anyone would be emailed, nothing changes until the inspector taps Send on the card this returns; voice cannot send it. On a draft it emails nobody.
inspection_draft_create
ChatGPTCreate a reviewable inspection draft for the signed-in inspector. The street address is the only required field: call on a clear request as soon as you have it, carrying every detail already given. This creates DRAFT/NOT_STARTED state and never sends invitations, email, SMS, or seller-agent notices; sendInvitations is refused on this surface. To make the draft a real inspection, finalize it with inspection_draft_finalize, then send invitations with inspection_invites_send under MCP's separate external-effect confirmation contract.
inspection_draft_finalize
ChatGPTTurn a saved inspection draft into a real scheduled inspection, exactly like finishing the create flow in the dashboard. Works only on a manual draft that has an address and a report template; drafts created from a client booking request must be accepted from Scheduling instead. Sends nothing; invitations are a separate step with inspection_invites_send. A request to finalize or schedule the draft authorizes it, as does a yes to your offer to finalize. A request to complete creation of the inspection includes relevant invitations: follow with inspection_invites_send unless the user explicitly excludes invitations. A request only to finalize sends nothing.
inspection_fee_apply
ChatGPTPut one of the inspector's saved fees on this inspection, by its name. Read inspection_services_list for the available fees. This changes the quoted price and emails every client attached to the job that it changed. When anyone would be emailed, nothing changes until the inspector taps Send on the card this returns; voice cannot send it. On a draft it emails nobody.
inspection_fee_remove
ChatGPTTake one fee off this inspection. The fee stays in the inspector's saved list and can be applied again; only this job loses it. Refused when the fee is not on the job, so read inspection_services_list first. This changes the quoted price and emails every client attached to the job that it changed. When anyone would be emailed, nothing changes until the inspector taps Send on the card this returns; voice cannot send it. On a draft it emails nobody.
inspection_get_context
ChatGPTRead the full authorized operational context for an inspection, including client, agent and contractor names/contact details, property address/details, schedule, canonical invoice totals and itemization, services, reports, units, assignments, and workflow status.
inspection_get_readiness
ChatGPTRead deterministic workflow readiness signals for the current inspection, including completion, review, upload, and reminder counts. analysisReview holds TWO separate review backlogs: unreviewedAiEdits (AI edits awaiting sign-off in the AI Review panel, what its Review badge shows, cleared by "Mark all reviewed") and openResultCards (analysis results awaiting confirm/decline on the AI Review results screen). Neither review clears the other. Never sum them and never call either "suggestions".
inspection_invites_send
ChatGPTEmail the inspection invitation to every client and real-estate agent already attached to this inspection who has not been invited yet, and text the ones with a phone number on file. Recipients come from authoritative inspection context, so never retype a name or email to send an invitation. Works only on a finalized inspection, not a draft. Call it when the user asks to create the inspection or send the invites, unless they exclude invitations. Nothing is sent until the inspector taps Send on the card it returns.
inspection_modifier_override
ChatGPTChange one automatic pricing adjustment on one service of this inspection: the square-footage band, the foundation surcharge, the year-built adjustment, or the travel-distance charge. Use action "set" with an amount to pin a figure, "remove" to charge nothing for it, or "reset" to hand it back to automatic pricing. Read inspection_services_list first: this can only move an adjustment the job already has, and adjustments come from the inspector's pricing settings, so it can never create one. This changes the quoted price and emails every client attached to the job that it changed. When anyone would be emailed, nothing changes until the inspector taps Send on the card this returns; voice cannot send it. On a draft it emails nobody.
inspection_schedule_update
ChatGPTChange the date, start time, end time, or walkthrough time of an existing inspection. Moving the date, start time, or walkthrough time immediately emails every client and agent attached to the inspection. Times are wallclock values in the property's local timezone; send only the fields that change. Call it for a clear request; when anyone would be emailed, nothing changes or sends until the inspector taps Send on the card it returns. On a draft it emails nobody.
inspection_service_add
ChatGPTAdd one service to an existing inspection by its catalog name. Read inspection_services_list first. Omit price to use the catalog price; pass price for a one-off line. This changes the quote and emails the attached clients on an issued inspection; requires the existing client-contact confirmation token. Drafts send nothing.
inspection_service_price_set
ChatGPTSet what one service costs on this inspection, overriding the catalog price for this job only. The inspector's catalog is unchanged. This changes the quoted price and emails every client attached to the job that it changed. When anyone would be emailed, nothing changes until the inspector taps Send on the card this returns; voice cannot send it. On a draft it emails nobody.
inspection_service_remove
ChatGPTTake one service off an inspection. Read inspection_services_list first: it says whether the service can be removed and whether removing it deletes report work already recorded under it. When it does, this refuses until you have told the inspector exactly that and they have agreed — only then call again with confirmReportDataDeletion true. This changes the quoted price and emails every client attached to the job that it changed. When anyone would be emailed, nothing changes until the inspector taps Send on the card this returns; voice cannot send it. On a draft it emails nobody.
inspection_services_list
ChatGPTRead the services, fees and discounts on an inspection with their prices and the current totals, plus what the inspector's own catalog offers. Call this before adding, removing or repricing anything: it is how you learn the exact service names, whether a service can be removed at all, and whether removing it would destroy report work.
inspection_set_working
ChatGPTSet the inspection this connection is working on. Later calls that do not name an inspection use it, and it stays set until you change it. Pass the inspection handle from inspections_search. Use this once the inspector names a job, rather than repeating the handle on every call.
inspection_unit_add
ChatGPTAdd one unit to an existing multi-unit inspection — a second dwelling, a guest house, an apartment in a duplex. Read inspection_units_list first so you do not repeat a name that already exists. Hometrace picks the unit number itself; never invent one. Square footage, year built, and foundation are per-unit PRICING inputs, so pass whatever the inspector gives you rather than letting the unit inherit the property figures. Adding a unit bills it on every per-unit service, creates its report, changes the quote, and can stale an agreement. A direct request to add the unit authorizes all of that: execute it and report the quote change. When you propose adding one, say it re-bills per-unit services and changes the quote; a yes executes it.
inspection_unit_remove
ChatGPTRemove one unit from this inspection, by the exact name returned by inspection_units_list. The unit's own report and inspection progress are deleted and cannot be brought back. Photos taken in that unit are KEPT in the inspection album — they stop being filed under a unit, that is all. Removing a unit changes the quoted price on every per-unit service and can make an already-signed agreement stale. A direct request naming the unit authorizes the removal: validate the exact unit immediately before deleting, execute, and report what was removed and the quote change. When the request is vague or the removal is your idea, state the permanent report/progress deletion and quote effect first; a yes executes it.
inspection_unit_update
ChatGPTRename one unit on an existing inspection, or set its square footage, year built, or foundation. Address the unit by the name the inspector uses; read inspection_units_list first when you are unsure. Square footage, year built, and foundation are per-unit PRICING inputs, so changing one re-quotes the job and can stale an agreement. A direct request with the changes authorizes it: execute and mention the re-quote when one of those values moved. When you propose the change, include that; a yes executes it. A rename alone does not change price.
inspection_units_list
ChatGPTList the units on a multi-unit inspection: number, name, square footage, year built, and foundation. Read this before adding or changing a unit — it is how you learn which names are taken and how many the inspection can still hold.
inspections_list
ChatGPTList the signed-in inspector's directly owned, non-deleted inspections as a deterministic cursor-paginated inventory. This is the complete-set read for account-wide questions and batch planning; inspections_search remains the small fuzzy lookup for finding one named job. Returns inspection handles and lifecycle, scheduling, archive, publication, payment, agreement, booking, recurring, sample, creation, and count metadata. Use the filters to define the intended set, request at most 100 records per page, and follow page.nextCursor until it is null before claiming you inspected the complete set. This tool only reads; it does not authorize or perform a batch action.
inspections_search
ChatGPTFind the signed-in inspector's inspections by street address, city, zip, or client/agent name. Misspellings are okay — when nothing matches exactly, the closest fuzzy matches are returned with a note. Returns inspection handles with address, status, date, and client names. Use this whenever the inspector refers to an inspection you do not have a handle for — never ask the inspector for one. Once the inspector has named the job, pass the handle to inspection_set_working so later calls do not have to repeat it.
inspector_get_account_context
ChatGPTRead the current signed-in inspector account and non-sensitive business, organization, profile, credential, notification, report, scheduling, service/pricing, and website settings. Payment instruments, provider identifiers, passwords, tokens, and security data are excluded.
inspector_memory_forget
ChatGPTForget one work preference previously saved by the current inspector. List memory first when the saved preference ID is unknown.
inspector_memory_list
ChatGPTList the current inspector’s bounded, inspectable memory. Direct-statement capture is enabled; hidden style inference and personality profiling are disabled.
integrations_list
ChatGPTRead integrations available to the signed-in inspector, connection status and exact setup paths. Call before answering integration questions; unavailable means not available yet for this account. Never claims credentials or connects anything.
photo_analysis_get_status
ChatGPTRead photo-analysis jobs, open analysis-review work, uploaded-photo counts, and unfiled-photo counts for the current inspection. analysisCoverage is the authoritative history for photos currently ready for analysis: EMPTY means explain that none are ready and do not start; NONE_ANALYZED means proceed normally; PARTIALLY_ANALYZED defaults to processing only notYetAnalyzedPhotos; ALL_ANALYZED means reuse saved results unless a rerun was explicitly requested. Photo coverage and successful jobs do not verify saved appliance fields. For missing or allegedly completed appliance details, read report_get_edit_context and distinguish saved values from blank fields before claiming completion. review.openResultCards counts result cards awaiting confirm/decline on the AI Review results screen; review.unreviewedAiEdits counts AI edits awaiting sign-off in the AI Review panel. They are separate backlogs over separate surfaces, so quote them separately. In an Analyze-button task, analysisIntent is the current server workflow state: read it after each question answer and follow READY, AWAITING_DECISION, or WAITING_FOR_SORT instead of stale earlier instructions.
photos_cover_set
ChatGPTSet or clear the inspection's cover photo, the photo on the report cover. The cover is shared by every report on the inspection. Pass mediaId to make that photo the cover, or remove: true for no cover. The photo is the one attached to the inspector's current message (its id line) or one they plainly identified. Never pick one from the album yourself; if it is unclear which photo they mean, ask. Say it is set only after this returns.
photos_get
ChatGPTLOOK at inspection photos: returns the images themselves, so you can see what is in them. Pass the handles from photos_list, and only entries it marked viewable. At most 8 photos per call, and best with one or two — a call naming more is refused outright, so ask for the ones you actually need, look at them, then call again for more. Anything not returned is listed under dropped with a reason; never describe a photo that is not in this result.
photos_list
ChatGPTList the inspection album's photos and videos: media handles, filenames, capture time, whether each is sorted, its current report placements (area/subsection/spec), its media type, and the handle of the SERVICE REPORT it belongs to (reportId). reportOutlines lists every report field's specId by area and subsection, for filing a photo ON its field with photos_sort. Albums span every service on the job, so when the inspector means one service, match on that report's handle yourself; this tool always lists the whole album. Only entries whose type is PHOTO can be displayed on MCP; photos_get will drop videos. This returns metadata only: it does not show you the media. To actually SEE photos, pass their handles to photos_get, up to four at a time and best with one or two; until you have, do not describe what a photo shows.
photos_sort
ChatGPTFile album photos or videos into this inspection report, each at the MOST SPECIFIC place it clearly belongs. When a report field (spec) matches what the item shows, file it ON that field: a driveway photo goes on the Driveway field, a water heater on the Water Heater field. That one placement both organizes the album and attaches the item to that field in this report. Use a subsection only when no field in it fits, an area only for media about the whole area, and leave an item unfiled when you cannot tell where it goes. A specId alone is enough: its subsection and area are filled in for you. Every field's specId is in photos_list's reportOutlines. If you know a field's name but not its id, send its subsectionId with fieldName and the field is found for you. The result's placements say where each item landed: say an item was attached to a field only when its placedOn is field, and that it went into the subsection's photo album otherwise. Media ids come from photos_list or the id line on the inspector's message. Never change the reusable report template. One call files into one service report. An item it cannot file (an id from elsewhere, a field on another service's report, media not in this album) is skipped, never the whole call: the result's skipped lists each with why. Tell the inspector which items were not filed and never say they were. When the inspector asks for media to be filed or sorted, choosing the placements is YOUR job, not theirs: look at the media, match what each item shows to the report's fields, subsections, and appliances, and file every item you can place. Never ask the inspector to name destinations you can determine by looking; "file these" with no destination is complete authorization on its own. A destination they do name, and any correction they make, always wins over your choice; when they name a section, still put each item on the field in it that matches. Inspectors say "section" loosely: when what they name matches a field's label, that field is the destination even if a section's name also contains the word ("the sub-panel section" is the Sub Panel Location field, not the Main & Subpanels section's album). A section's photo album is only for items no field in it fits. When they name one and an item plainly does not belong there, do not force it in and do not quietly file it somewhere else: leave that item unfiled, say which one it is and what it shows, and ask whether they want it where it belongs or somewhere else. File the rest without waiting. This does not put a photo on a note, limitation, or deficiency card: pass mediaIds to report_finding_create (a new item) or report_finding_update (an existing one) for that, and do not also file that photo on the field the item sits under, or the report shows it twice.
report_appliance_fields_set
ChatGPTWrite appliance data-plate details (make, model, serial number, manufacture date, capacity, fuel or heat type) into the appliance's own fields on the current authorized draft report. This is how plate data gets onto the report, whether the inspector dictated it or you read it off a plate photo you can see: NEVER read the values back and ask the inspector to type or paste them in, never ask them to read out what the photo shows, and never park them in a note instead. Read report edit context first to get the applianceId and to see which fields exist, what a select field allows, and which are already filled. Send only the fields you actually read; anything you omit keeps its current value. This setter does not clear fields; use Remove in AI Review to undo AI edits or restore empty values. A field the inspector already filled with something different is skipped and reported back rather than overwritten; read the conflict out and ask them which is right. Unknown labels and disallowed option values come back in the result too; fix them and call again. Write the serial number exactly as it is printed and NEVER decode it yourself: this tool decodes the serial deterministically from the brand's published format, and when the decode is certain it fills an empty manufacture-date field. A date you read off the plate is written as you sent it; when the decode disagrees, the result says so, and you tell the inspector both dates so they can check the plate. Report the date the way the result states it, and never state one the result did not confirm.
report_area_create
ChatGPTAdd one new area to the current authorized draft report — an area is the TOP level of the report tree ("Interior", "Roof", "Detached Garage"), the thing that holds sections, which in turn hold fields. Use it only when the property has a whole part the report has no place for at all; if the report already has an area that could hold it, add a section there with report_subsection_create instead. The new area arrives EMPTY: sections go in afterwards with report_subsection_create, one call each, and fields inside those with report_spec_create. It is added to THIS inspection only and the inspector's template is untouched, so it changes nothing about their next job. Read report edit context first and check the area is not already there under a different name — area names are unique on a report, so a name that is taken is rejected. Do not use it to rename or reorganize an existing area.
report_area_delete
ChatGPTRemove one whole area, and everything anywhere inside it, from the current authorized draft report. This is the widest report change: the area, all sections, all fields and their answers/comments, all findings, and all owned appliances with saved values are permanently deleted. Re-adding the area creates an empty one. A direct request naming the area authorizes it; when the removal is your idea or the request is vague, state this whole-area scope, and a yes executes it. To remove one section use report_subsection_delete; one field use report_spec_delete; one item use report_finding_delete. Read authoritative report context immediately before calling, disclose the current counts of sections, fields, findings, and appliances, and make sure the whole-area scope has been said out loud, by them or by you, before deleting. Pass the exact displayed name so an id/name mismatch refuses safely. Execute the completed request without a separate confirmation and report exactly what the result says was removed.
report_finding_create
ChatGPTCreate one general note, limitation, or deficiency on the current authorized draft report. Take the subsection or field from the report map, or read report edit context when it is not there. A subsection with no specs is a NORMAL target: pass subsectionId alone and the item lands in that subsection's General section, which is exactly where a manual subsection-level note goes. Never create a spec in order to have somewhere to put a finding. Always send the whole item on the first call (itemName, type, description, and severity only when the inspector rated it), never an empty call and no report_template_notes lookup first: in a section with no templated notes it is created at once. Templates first: in a section that has the inspector's templated notes, a call with no templateNoteId creates nothing and returns that section's notes; call again with the templateNoteId whose meaning matches what the inspector described, or, only when none fits, with noTemplateFits: true and your own wording, without asking the inspector. Never mention templates, templated or saved notes, or where the wording came from in your reply; just say what you added. An itemName that is a note's exact title uses that note. When the inspector sent or captured a photo of what this item describes, put it on the item's card with mediaIds in this same call, without asking; never tell them to attach it by hand. One card per thing in a section: when the inspector names an item that section already has, even as "add a defect for X", add to that card with report_finding_update instead of creating another.
report_finding_delete
ChatGPTRemove one note, limitation, or deficiency from the current authorized draft report. This is the ONLY way to remove a report item — saying you removed it, or writing a note that says it was removed, does not remove anything. The item is taken off the report; its photos are kept. Use it only when the inspector asks for that item to come off, and never as a way to rewrite one: to change wording or severity, revise the item instead. To get the item's ID, read report edit context narrowed to the subsection it sits in, or to the field it is about, and use the findingId it returns.
report_finding_update
ChatGPTRevise one note, limitation, or deficiency that already exists on the current authorized draft report. This is the ONLY way to change a report item — describing an edit in your reply does not change anything. Send only the fields you are changing: anything you omit keeps its current value, so never resend the whole item to change one part of it. To get the item's ID, read report edit context narrowed to the subsection it sits in, or to the field it is about, and use the findingId it returns. To put a photo on an item that already exists, pass mediaIds, alone or with other changes.
report_get_edit_context
ChatGPTRead one bounded, paginated, authorized edit slice for a Hometrace report: areas, subsections, specs, current answers, option IDs, custom responses, saved comment options, ordered comments, each subsection's appliances with their field labels, types, options, and current values, and the report items already on each subsection (notes, limitations, and deficiencies) with the findingId needed to revise or remove one. This is the ONLY way to learn the ID of an item you did not create yourself in this turn. Read it before reporting appliance data-plate details so you can see which fields exist and which are already filled, and before changing or removing any existing item. Narrowed to a subsectionId or specId, it also returns templateNotes: the inspector's templated notes for that target, the same list report_template_notes gives, so there is no need to call that tool too. A read with no subsectionId, specId, or query is a summary (detail "summary"): each field lists only its selected options with an optionCount, and no saved comment options; narrow before choosing an option or saved comment. Report items that belong to no subsection come back in reportLevelFindings, and only on a read that narrows to nothing. Narrow by target or query before choosing report edit IDs. A read already made this turn is still current for its ids. Follow pagination.nextCursor when needed and never guess omitted targets.
report_page_create
ChatGPTAdd one TEXT custom page to the inspector's report pages so future Hometrace branded (and four-point) generated PDFs include it. Call report_pages_list first. If a page already has this title, this REFUSES rather than updating; use report_page_update. Type is always TEXT; there is no PDF upload on this tool. Default position is AFTER_COVER (after the cover sheet, before the report body), which is where a state-required summary disclaimer belongs. On four-point forms that placement is remapped to the end of the form. Content may be plain text or the HTML subset the PDF renderer understands (p, h2, h3, lists, bold/italic/underline, links). Plain text is wrapped the same way Settings > Report wraps it. 10,000 characters of visible text maximum. This changes the inspector's custom pages for future Hometrace branded (and four-point) generated PDFs. Already-cached PDFs and already-built reports are unchanged until regenerated. No client is emailed. This is not a finding, not the overview summary, and not a signable agreement. To put wording on one inspection, use the report write tools. To edit a pre-inspection agreement, use Settings > Documents, not these tools. TREC native PDFs do not render TEXT custom pages; uploaded PDF-type pages still splice. On four-point forms, AFTER_COVER is remapped to the end of the form.
report_page_update
ChatGPTChange one TEXT custom page: rename it, replace or append its body, move it between after-cover and end-of-report, or enable/disable it. Address it by the title the inspector uses. Call report_pages_list first. If the named page is an uploaded PDF, this REFUSES; do not convert it, and do not overwrite it. Create a separate TEXT page instead if the inspector wants the wording in addition to the PDF. mode=append concatenates onto the existing body without dropping it; default is replace. 10,000 characters of visible text maximum on the resulting body. This changes the inspector's custom pages for future Hometrace branded (and four-point) generated PDFs. Already-cached PDFs and already-built reports are unchanged until regenerated. No client is emailed. This is not a finding, not the overview summary, and not a signable agreement. To put wording on one inspection, use the report write tools. To edit a pre-inspection agreement, use Settings > Documents, not these tools. TREC native PDFs do not render TEXT custom pages; uploaded PDF-type pages still splice. On four-point forms, AFTER_COVER is remapped to the end of the form.
report_pages_list
ChatGPTList the inspector's custom report pages: the extra sheets included in Hometrace branded (and four-point) generated PDFs (standards of practice, a state summary disclaimer, terms). TEXT pages include their full content. PDF pages return type, title, position, order, isEnabled, and pageCount, with content null; they are uploaded files and cannot be edited here. Call this before creating or updating a page so you use the exact title and do not duplicate one. This is not the overview summary, not a finding, and not a signable agreement. TREC native PDFs do not render TEXT custom pages.
report_spec_answer_set
ChatGPTSet, add to, or clear one structured answer on a report field. A choice field takes option LABELS (selectedOptionLabels) or option ids; selection add or remove changes only those options. expectedUpdatedAt is optional: without it the write fills an empty field, repeats its value, adds or removes options, or replaces an unreviewed AI answer, and refuses to replace any other answer, returning it with the updatedAt to send if the inspector asked for the change. With it, a stale value is refused with SPEC_CONFLICT; then re-read the field.
report_spec_create
ChatGPTAdd one inspection-only checkbox, unit-value, comment, or yes/no field to an authorized subsection of the current draft report. This never changes the inspector template. Execute a clear user-authored request once the subsection, field label, and complete field definition are known.
report_spec_custom_response_set
ChatGPTSet, replace, or clear the custom inspector response attached to one authorized report spec. expectedUpdatedAt is optional: without it the write fills an empty response or repeats it, and refuses to replace a different one, returning it with the updatedAt to send if the inspector asked for the change. NEVER use this for a limitation: anything that limited or prevented inspection (stored items, weather, inaccessibility, not visible) is a finding: create it with report_finding_create, type LIMITATION, specId set to this field.
report_spec_delete
ChatGPTRemove one field from the current authorized draft report. This is the ONLY way to take a field off the report — saying you removed it, or writing a note that says it was removed, does not remove anything. Removing a field also permanently removes the comments on it and its photo links; notes, limitations, and deficiencies stay on the report, just no longer attached to that field. A direct request naming the field authorizes it; when the removal is your idea or the request is vague, state this full deletion scope, and a yes executes it. To change a field's answer use report_spec_answer_set, and to clear its custom response use report_spec_custom_response_set with a null note. Read authoritative report context immediately before calling, validate the exact field ID, execute without a separate confirmation, and report exactly what the result says was removed.
report_subsection_complete_set
ChatGPTMark one existing section on the current authorized draft report complete or incomplete. This is the ONLY way to change a section's completion state. Read report edit context first, use the exact section id and name it returns, and pass completed true to mark it done or false to reopen it. A TREC section must already have an I/NI/NP/D inspection status before it can be marked complete. Report success only after this tool returns the section and its confirmed completion state.
report_subsection_create
ChatGPTAdd one new section to an area of the current authorized draft report — a subsection such as "Guest Bathroom" or "Detached Garage" under an area such as "Interior". Use it when the inspector found something the report has no place for; adding the fields inside it afterwards is report_spec_create, one call per field. The section is added to THIS inspection only and the inspector's template is untouched, so it changes nothing about their next job. Read report edit context first to get the area ID and to check the section is not already there under a different name. Do not use it to rename or reorganize an existing section.
report_subsection_delete
ChatGPTRemove one whole section, and everything inside it, from the current authorized draft report. The section, all fields and their answers/comments, all findings, and all owned appliances with saved values are permanently deleted. Re-adding the section creates an empty one. A direct request naming the section authorizes it; when the removal is your idea or the request is vague, state this whole-section scope, and a yes executes it. To remove one field use report_spec_delete; one item use report_finding_delete; to blank an answer use report_spec_answer_set. Read authoritative report context immediately before calling, validate the exact id and displayed name, and make sure the whole-section scope has been said out loud, by them or by you, before deleting. Execute the completed request without a separate confirmation and report exactly what the result says was removed.
report_subsection_inid_set
ChatGPTSet the Texas TREC I/NI/NP/D boxes on one existing section of the current authorized draft TREC report. The boxes combine (e.g. INSPECTED + DEFICIENT); the list replaces the current boxes and an empty list clears them. Read report edit context first and use the exact section id and name it returns. Report the boxes only after this tool returns them.
report_template_notes
ChatGPTList the inspector's TEMPLATED NOTES for the current report: canned findings/notes saved on their report template, in their own preferred wording. Pass specId when the note belongs under a report field so both exact-spec and subsection notes are included; otherwise pass subsectionId or areaId. Omit all three only when the inspector asks to browse their whole template library. PREFER a matching templated note over freeform writing. You do not need this before report_finding_create: send it the whole item and it returns the section's templated notes only when the section has any. When one fits what the inspector described, create the finding from it with report_finding_create's templateNoteId instead of retyping its wording. Never mention templates, templated or saved notes, or where the wording came from in your reply; just say what you added. When none of the target section's notes fits, write the item in your own wording (report_finding_create with noTemplateFits: true) without searching the whole library.
template_area_add
ChatGPTAdd one area to a report template — a top-level section like "Detached Garage" or "Outbuilding". A General subsection is created inside it automatically, which is where a note with no specific target lands. This is refused on a standardized form, whose areas have to match the form. A template is the FORM, not a report. Reports that already exist keep the structure and wording they were built with; this only changes what future reports start from.
template_get
ChatGPTRead one report template whole: its areas, their subsections, the fields in each subsection with their answer type, and how many templated notes are offered in each place. Narrow it to one area with areaName to also see each field's answer options. Read this before changing a template's structure so you name areas, subsections and fields exactly as they are. The notes themselves come from template_notes_list.
template_import_chunk_put
ChatGPTStore one numbered chunk of an import (areas and/or template-wide notes, shaped as template_import_guide describes). Send each area and item once with a stable key; repeating a key merges. Number chunks 1, 2, 3; re-sending a number replaces it. About 60 notes per chunk.
template_import_commit
ChatGPTCreate the new template from a previewed import, in one step, after the inspector agrees. Never changes an existing template. Calling it again returns the same template.
template_import_guide
ChatGPTRead FIRST when the inspector wants to bring a template over from HomeGauge, Spectora, ISN, Home Inspector Pro, Word, PDF or a spreadsheet. Returns the import format, how each system maps onto Hometrace (areas, items, Styles & Materials, notes, note types and severity), limits, and the step-by-step workflow.
template_import_preview
ChatGPTDry run of an import: contents, merges, likely duplicates and gaps, classifier calls to check, warnings, errors that block commit. expect = your source counts. Changes nothing.
template_import_start
ChatGPTOpen an import that will become a NEW template; nothing is created yet. Returns an importId. Starting again with the same name resumes that import. Follow template_import_guide.
template_note_add
ChatGPTSave one templated note to a report template: the inspector's own wording, kept so it can be reused instead of retyped. This is the write half of report_template_notes, which is how the saved wording comes back later. Give it a short title and the full wording, and name the area or subsection it belongs under so it is offered in the right place. Allowed on a standardized form, because wording is theirs even when the structure is not. A template is the FORM, not a report. Reports that already exist keep the structure and wording they were built with; this only changes what future reports start from.
template_note_duplicates_find
ChatGPTScan every templated note on one report template and return the groups that say the same thing: identical_wording (same words, ignoring case and punctuation), similar_wording (mostly the same words), and same_title (one title, different wording, often an old and a revised version that the inspector may want to keep both of). Each group lists its notes in full with where each is offered and how often it was used, and names a suggestedKeepNoteId (most used, then most placements, then longest). Changes nothing. Removing or merging duplicates is done in the Hometrace app chat or the template editor.
template_note_update
ChatGPTChange one templated note the inspector already saved: its title, its wording, or its severity. Name it by the templateNoteId from template_notes_list, template_note_duplicates_find or report_template_notes, never by typing an id the inspector gave you, and send only the fields that change. Findings already written from this note keep the wording they were written with; this changes what future reports offer. Allowed on a standardized form, because wording is theirs even when the structure is not. A template is the FORM, not a report. Reports that already exist keep the structure and wording they were built with; this only changes what future reports start from.
template_notes_list
ChatGPTRead the templated notes saved on one report template, a page of up to 100 at a time: each note's templateNoteId, title, full wording, kind, severity, where it is offered (areas, subsections, fields; empty means template-wide), and how many report items were written from it. Narrow with areaName (and subsectionName) or a query. Needs no inspection. Page on with nextCursor until it is null. To find duplicates, call template_note_duplicates_find instead of reading every page.
template_rename
ChatGPTRename one of the inspector's report templates, or change its description. This works even on a standardized form whose structure is locked: the name is theirs, the shape is not. A template is the FORM, not a report. Reports that already exist keep the structure and wording they were built with; this only changes what future reports start from.
template_subsection_add
ChatGPTAdd one subsection under an existing area of a report template — for example "Flashing" under "Roof". Name the area it belongs in. This is refused on a standardized form, whose structure has to match the form. A template is the FORM, not a report. Reports that already exist keep the structure and wording they were built with; this only changes what future reports start from.
templates_list
ChatGPTList the inspector's report templates: name, description, whether each is their default, whether its structure is locked, and how many areas it has. Read this before changing a template so you use the name they actually gave it, and so you know whether its structure can be edited at all. For what is inside one template, use template_get; for its notes, template_notes_list.