MCP App Store
by KeenEthics
ShopBooster icon

ShopBooster MCP Server for ChatGPT

By ShopBooster

Overview

ShopBooster is the CRM for window tint, PPF, ceramic coating and detailing shops. Connect it to see your leads, conversations, schedule, quotes and invoices, and get work done without opening the app. Ask for the exact price for a customer's car, straight from your shop's price list. Draft a quote, find an open time and book the job. See what needs attention today, catch up on a customer's thread, suggest a reply for your team, or add a note or tag. ShopBooster only quotes prices from your own list and never invents discounts. Anything that reaches a customer (a text, a quote, an invoice link, a booking confirmation) is previewed first and only goes out after you say yes. Opted-out and blocked customers are never texted, and every action is logged as "Shop Bot". The shop owner chooses on the sign-in screen what the AI may do, and can switch any connection off in Settings → Integrations. Requires a ShopBooster account (owner role to connect). US shops.

How to connect

  1. ChatGPT: Open the ShopBooster listing in the ChatGPT app directory and choose Connect. ChatGPT listing ShopBooster exposes 48 tools here.

Tools

add_customer_note

ChatGPT
Adds an internal note to a customer's Notes timeline (seen on the contact, their leads, appointments and Messages; not by the customer). Signed "Shop Bot (approved by ...)". approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

add_lead_note

ChatGPT
Adds an internal note to the lead's customer Notes timeline (seen on the lead, the contact, their appointments and Messages; not by the customer). Signed "Shop Bot (approved by ...)". approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

approve_time_correction

ChatGPT
Approves a tech's time-correction request. It changes recorded hours, which payroll reads. Recorded as "Shop Bot (approved by ...)". approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

archive_conversation

ChatGPT
Moves a conversation out of the main inbox into Archived. A new customer message brings it back. Nothing reaches the customer. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

book_appointment

ChatGPT
Books a confirmed job at starts_at (ISO time with offset, or "YYYY-MM-DD HH:MM" in the shop's time zone). A taken or closed slot is refused with slot_unavailable and alternatives. Prices come from the shop's price list for the vehicle type: a missing price is refused with price_missing (nothing is created), and a price sent in the request is ignored and reported as a price_ignored warning with the price that was used. An option package takes exactly one option. The shop's booking confirmation email and text go out to the customer (the text is skipped for opted-out or blocked customers). approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

cancel_appointment

ChatGPT
Cancels a job from the work-status menu, which frees the slot. ShopBooster sends the customer no cancellation message (customer_notified: false). Jobs in progress or done are refused. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

check_message

ChatGPT
Runs every rule a send would (card or ID details, cussing, opt-out, blocked, app links, sending hours, daily cap) and the warnings (a price for an unknown vehicle, a price not on the list, a closed or booked time) without sending anything. Returns check_id, ok, blocks, warnings, expires_at (15 minutes) and show_the_owner (one plain line summing up the check). While the shop has "Bots must check before sending" on (the default), a send needs the check_id of a check of exactly the same text and photos, made at least 5 seconds earlier and not expired; a check with warnings also needs warnings_approved: true on the send, and a check with blocks can't be used. Each check_id works for one send.

create_contact

ChatGPT
Adds a contact, or returns the existing one (matched: true) when a contact in this shop already has that email or phone. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

create_lead

ChatGPT
Adds a lead (New Lead) for contact_id or an inline contact (matched by phone or email first). The app's new-lead flow runs. The shop's new-lead auto text is not sent for leads added here (new_lead_auto_text.scheduled is false), so no text reaches the customer. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

decline_time_correction

ChatGPT
Declines a tech's time-correction request (their hours stay as recorded). approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

draft_invoice

ChatGPT
Creates a draft invoice: either a booked job's balance (appointment_id) or packages from the price list (packages + vehicle_type). Draft: nothing is charged or sent. Amounts come from ShopBooster, not the request. A discount goes in only as discount_id, one of the shop's saved discounts (any other id is refused with discount_not_allowed); ShopBooster works out the amount. The vehicle's year, make and model are needed (passed, or on the linked lead, quote, job or customer), else the call is refused with vehicle_required. customer_url is the customer's pay page (the one invoice link a customer can open); app_url is the invoice in ShopBooster, for the shop's staff. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

draft_proposal

ChatGPT
Creates a draft proposal (quote) in ShopBooster. Draft: nothing is sent to the customer. Prices come from the shop's price list for the vehicle type: a missing price is refused with price_missing (nothing is created), and a price sent in the request is ignored and reported as a price_ignored warning with the price that was used. A discount goes in only as discount_id, one of the shop's saved discounts (any other id is refused with discount_not_allowed); ShopBooster works out the amount. The deposit is off unless deposit is passed. Returns the proposal with app_url (the draft in ShopBooster, for the shop's staff) and customer_url (the customer's approval page, the one proposal link a customer can open). approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

duplicate_proposal

ChatGPT
Copies a proposal as a new draft for the same customer, e.g. the same quote for their other car. Draft: nothing is sent to the customer. vehicle changes the car (each field passed replaces the old one) and packages changes what's quoted; otherwise the same packages and options are copied. Everything is re-priced from today's price list for the vehicle, not from the old proposal or the request: size_required and price_missing apply. The copy is linked to no lead unless lead_id is passed; then it becomes the lead's one linked proposal and the lead's earlier one is unlinked (kept, not deleted; the preview warns). Returns app_url (the draft in ShopBooster, for the shop's staff) and customer_url (the customer's approval page). approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

find_customers

ChatGPT
Searches the shop's customers by name, phone number or email and returns their ids. Matching: name is case-insensitive and every word has to appear ("jane doe" finds "Jane Q. Doe"); a phone in any format ("(805) 555-0100", "+18055550100") is compared on its last 10 digits across all of the contact's phones; email is case-insensitive, exact or partial. Results come best match first (match.field and match.score), then most recent activity, with phones, emails, vehicles, tags, latest_lead, latest_proposal, opted_out, blocked, last_contacted_at and url (the contact in ShopBooster, for the shop's staff). Several results are several different people. tag (e.g. "VIP") keeps only customers with that tag; tag without search lists that tag's customers, newest first. ids (up to 100) returns those customers in the order given instead of a search, with match null and missing_ids for ids not found. Texts to opted_out or blocked customers are refused. Adding a contact whose phone or email already belongs to a customer here returns that customer (matched: true) instead of a duplicate.

find_open_times

ChatGPT
Open start times between two shop-local dates (14 days at most), from the shop's hours, lunch, bays, booked jobs and techs. duration (minutes), or package_ids with vehicle_type (and size) for the shop's own job times. Times only, no prices, so year, make and model aren't needed here. Read-only; doesn't count toward the daily cap. Booking a time that is taken or closed is refused with slot_unavailable.

get_action_log

ChatGPT
What assistants did in this shop: every text sent, action run and connection change, newest first, with action, ids, result (ok, matched or the refusal code), http_status, approved_by, approved_at, draft_id and which key. Refused attempts are included. Suggested-reply rows (drafts.create, drafts.sent, drafts.dismissed, drafts.expired, drafts.outdated) also carry suggested_text, sent_text and edited (texts_withheld when the owner keeps edits private). Read-only. Filters: since, action (e.g. "leads.move_stage"), result (e.g. "quiet_hours"), request_id (a response's X-Request-Id). Each entry has request_id. Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time. next_cursor, passed back as cursor with the same filters and order, continues the list; next_cursor null means the list is complete.

get_conversation_messages

ChatGPT
One thread's messages and calls, oldest first, plus its state (opted_out, blocked, awaiting_reply). Photos are signed links valid for 24 hours. A missed call can carry a voicemail with its transcript and a recording_url (signed, 24 hours). next_before pages further back. context is the thread's customer (null customer_id when the thread isn't linked to one): their newest leads (stage, vehicle asked about, services requested, vehicles_from_proposals, auto_text and texting_summary), proposals, invoices (status sent/partly_paid/paid, balance_due, customer_url), appointments (starts_at in the shop's time zone, status booked/rescheduled/cancelled/completed/in_progress, tech) and the latest 5 staff notes, 10 of each at most, with counts. A proposal has status (draft/sent/opened/approved/declined/expired, the normalized value; app_status is the raw app value), quoted_vehicle, items (every package and option with the prices, durations and warranty as quoted on that proposal), open_count, total {from, up_to}, deposit and customer_url. On items, selected means the customer picked it; on a proposal that isn't approved, selected false means offered, not chosen. What a customer actually bought is shown by an approved proposal, a booked or completed appointment or a paid invoice. include: "events" adds CRM events (kind: "event": lead created or moved, proposal sent/opened/approved/declined, deposit paid, invoice sent/paid, appointment booked/rescheduled/cancelled/completed) between the messages by time. auto_text is the shop's automatic text for a website-form lead (status scheduled, sending, sent, failed, skipped, cancelled_team_texted or off; at; reason; null for other leads); texting_summary is one line on whether the customer was texted and what is on its way. scheduled_texts lists texts queued to go out to this customer (type new_lead_auto_text or scheduled_text, status, send_at). customer_url is a page the customer can open; app_url and url open ShopBooster for the shop's staff, and texts containing them are refused (internal_link). A conversation_id that isn't this shop's (another shop's, or one from another ShopBooster such as staging) is not_found with error.reason not_in_this_shop.

get_customer

ChatGPT
Everything the shop knows about one customer, by id: phones, emails, vehicles (vehicle_details: year, make, model, vehicle_type, size and source: appointment, proposal or lead), appointments, proposals, invoices, payments (brand and last 4 at most), notes and leads (id, lead_id, stage, vehicle, services requested, vehicles_from_proposals, auto_text and texting_summary, newest first), plus opted_out and blocked (texts to either are refused). conversation is their newest thread (source and id; null when they have none yet), last_inbound_at and last_outbound_at say when they last wrote in and the shop last wrote out, and scheduled_texts lists texts queued to go out to them with send_at (a lead that came in after hours gets the automatic text when sending hours start; its thread exists only once it goes out). customer.tags are the shop's own labels on them (e.g. "VIP", "Fleet"). Proposals have status (the app's raw value), normalized_status (expired once past its expiration date), quoted_vehicle and items (every package and option as quoted on that proposal, with selected and warranty_terms). On a proposal that isn't approved, selected false means offered, not chosen; what a customer actually bought is shown by an approved proposal, a booked or completed appointment or a paid invoice. customer_url is a page the customer can open; url and app_url open ShopBooster for the shop's staff, and texts containing them are refused (internal_link).

get_lead

ChatGPT
One lead with the same fields as the lead list: stage and full stage_history, outcome with outcome_at and outcome_value, lost_reason and lost_note, customer, vehicle, services, source and attribution, linked proposals (with customer_url) and appointment_ids, conversation, last_inbound_at and last_outbound_at. auto_text is the shop's automatic text for website form leads (status scheduled, sending, sent, failed, skipped, cancelled_team_texted or off; at; reason; null for other leads), scheduled_texts lists texts queued to go out to the customer ({type: new_lead_auto_text or scheduled_text, status: scheduled or sending, send_at, conversation}; a lead that came in after hours is texted when sending hours start, and its thread only exists once that text goes out) and texting_summary says in one line whether the shop has texted them and what is on its way. Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time. Outside card, the links returned are customer_url (a page the customer can open) and conversation.url (the thread in ShopBooster, for the shop's staff; texts containing it are refused with internal_link).

get_message_status

ChatGPT
Where one outbound message stands (any sender: Shop Bot, teammates, automations): status (texts: queued, sent, delivered, undelivered, failed; Facebook/Instagram: sent, failed, then delivered or seen), is_final, delivered_confirmed, history (each status with its time) and, when it didn't go through, error {code, reason, message} with a plain-English reason. delivered_confirmed is true only when the carrier or Meta confirmed it (a text: delivered; Facebook/Instagram: seen, or delivered on Messenger); queued and sent mean sent but not confirmed delivered. A text can stay sent for good when the carrier does not report back. is_final true means the status won't change again. The id and source come from a send's delivery.status_url, a webhook or a conversation's messages.

get_pending_action

ChatGPT
Where one held action stands, by the pending_action.id from the 202: status (pending, approved, edited, dismissed, expired or outdated), who acted and when, what was asked (request), what was used if a teammate edited it (changes, final) and the ids it made (result_ids). The same changes arrive as action.* webhooks. Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time.

get_price

ChatGPT
Returns the shop's exact price for one vehicle from its own price list. ShopBooster finds the vehicle in its vehicle catalogue and picks the vehicle type and size band (small, medium, large) the same way the shop's staff do. With package_ids and/or option_ids (any number; every active one fits) it returns, per package and option, the price, duration_minutes and warranty, plus vehicle_resolved {type, size}, resolved_by, how_resolved, confidence (exact, high, given or none), included (plain lines of what is covered; includes lists the smaller packages a combined package already covers) and total. total is null when the result holds alternatives (options of one package, such as two films) rather than one combined job. Without package_ids and option_ids it returns every package the shop prices for that vehicle (scope: all_packages, grouped in services, at most 60; service_ids narrows it): separate choices with total null. needs lists what is missing before there is a price: ["size"] or ["vehicle_type"] (the size or type can't be told from the vehicle; bands shows each size's price, or the shop's prices on each vehicle type, and the result has no single price), ["trim"] (trims differ in size) or ["price"] (the shop has no price for an item on this vehicle). A call without year, make and model (and no customer_id with one vehicle on file) is refused with vehicle_required. Other refusals list error.problems [{field, code, message, allowed, example}]. A warning "already included" means two chosen packages overlap. Discounts are not applied to these prices; available_discounts lists the saved discounts this customer and job qualify for.

get_proposal

ChatGPT
One proposal with the same fields as the proposal list (status, packages and options with their quoted prices, total, deposit, opens, expiry, customer, vehicle, customer_url), plus card: the quote card (apps that show cards draw it; card.markdown is the same card as text). Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time.

get_shop_info

ChatGPT
Returns the shop's details, opening hours, every service/package/option with prices_by_vehicle (and sizes when the price depends on vehicle size), warranty text, FAQ, booking settings, deposits, its saved discounts, example replies, reply_tone, reply_rules, reply_style.examples (the shop's own past replies, scrubbed), reply_style.corrections (how the shop edited AI suggestions, scrubbed) and bot_send_hours (the hours Shop Bot texts may go out). reply_tone, reply_rules and reply_style describe how this shop writes to customers. Prices here are the shop's list: price_from is the lowest price on the list, not a quote; exact prices for one vehicle come from the price lookup. The discounts list is every discount the shop offers (empty means none). A warranty of null means the shop states no warranty.

get_webhook_status

ChatGPT
How each of this shop's webhook endpoints is doing: url_host (not the full URL), events, active, status (healthy, failing or disabled), last_success_at, last_failure_at, consecutive_failures, failing_since and alerted (the owner was already told). Events sent while an endpoint was failing stay in the event replay. The owner gets one alert after about 30 minutes or 30 failed attempts in a row. Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time.

list_appointments

ChatGPT
The schedule: jobs overlapping from..to (default today plus 14 days, at most 62 days), soonest first. Each has status (confirmed, pending or canceled), work_status, starts_at, ends_at, services, techs, customer, vehicle, lead_id, proposal_id, and conflicts: true when another confirmed job has the same tech booked at the same time (conflicting_appointment_ids; overlapping_appointment_ids lists every overlapping job). Filters: from, to, tech_id, status, tag (a tag the customer has), updated_since (with only updated_since, every changed job whenever it is). Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time. next_cursor, passed back as cursor with the same filters and order, continues the list; next_cursor null means the list is complete. ids (up to 100 known ids, e.g. from webhooks or other lists) returns those records in one call, in the order given, with missing_ids for any not found.

list_conversations

ChatGPT
SMS, Facebook and Instagram threads ordered by last activity (oldest first by default; sort: "desc" for newest first; 50 per page). Each has state: awaiting_reply, last_outbound_by (user, automation or bot), assigned_to, muted, snoozed, archived, blocked and opted_out. Texts to opted_out or blocked threads, and to archived ones, are refused. muted and snoozed mean the shop set the thread aside. awaiting_reply: true lists only threads waiting on the shop (with older_than, e.g. "2h", only those waiting at least that long). scheduled_texts lists texts queued to go out in the thread (or to its customer, like a new lead's automatic text waiting for sending hours): {type new_lead_auto_text or scheduled_text, status scheduled or sending, send_at}. next_cursor continues the list.

list_drafts

ChatGPT
Shop Bot's suggested replies and their outcomes, latest change first: status (active, sent, dismissed, expired, outdated or replaced), suggested_text, sent_text (what actually went out), edited (a teammate changed it before sending), acted_by (the teammate, the customer or the system) and acted_at, outdated_reason (team_replied, customer_replied or context_changed), reason (why it was suggested), customer and conversation. When the owner keeps edits private (their "Learn from your edits" setting is off, or they removed that edit), sent_text is null and texts_withheld says why. status active means the suggestion still waits for a teammate. Filters: status, since (changed at or after), conversation ("sms:4182"). Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time. next_cursor, passed back as cursor with the same filters and order, continues the list; next_cursor null means the list is complete.

list_events

ChatGPT
The events this shop's webhook endpoints were sent, oldest first, each in the exact webhook envelope (id, type, created_at, occurred_at, sequence, data), for catching up on missed webhooks. since is the id of the last event handled (or an ISO time). Events are kept for retention_days (complete: false means since is older than that). An event id is the same whether it arrives here or by webhook. A shop with no webhook endpoint has no events. next_cursor, passed back as cursor with the same filters and order, continues the list; next_cursor null means the list is complete.

list_invoices

ChatGPT
Invoices, newest change first. Each has status (draft, unpaid, partly_paid or paid), amount, amount_paid, balance_due, due_date, overdue, is_deposit, customer, appointment_id, proposal_id, view_count, last_viewed_at and customer_url (the page where the customer views and pays; null for drafts). conversation is the customer's latest thread (source, id, path); last_inbound_at and last_outbound_at say when the customer last wrote in and the shop last wrote out. status=void returns nothing: ShopBooster has no voided invoices. Filters: status, overdue (true: a balance past the due date), since (created at or after), tag (a tag the customer has), updated_since; sort: balance (biggest first with order desc) instead of the newest change. Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time. next_cursor, passed back as cursor with the same filters and order, continues the list; next_cursor null means the list is complete. ids (up to 100 known ids, e.g. from webhooks or other lists) returns those records in one call, in the order given, with missing_ids for any not found.

list_leads

ChatGPT
Leads in the sales pipeline (people who asked the shop for a quote or a service), newest change first; order=asc for oldest first. Each lead has its stage and stage_history, outcome (open, replied, quoted, booked, deposit_paid, completed or lost), the customer and vehicle, the services they asked about, their message, where it came from (source), returning and previous_leads when the customer asked before, and linked proposals and appointments. conversation is the customer's latest thread (source, id, path); last_inbound_at and last_outbound_at say when the customer last wrote in and the shop last wrote out. auto_text is the shop's automatic text for website form leads (status scheduled, sending, sent, failed, skipped, cancelled_team_texted or off; at; reason; null for other leads), scheduled_texts lists texts queued to go out to the customer ({type: new_lead_auto_text or scheduled_text, status: scheduled or sending, send_at, conversation}; a lead that came in after hours is texted when sending hours start, and its thread only exists once that text goes out) and texting_summary says in one line whether the shop has texted them and what is on its way. Filters: stage (e.g. "New Lead", any case), source (e.g. "google"), outcome, since (created at or after), tag (a tag the customer has), updated_since. With outcome a page can hold fewer than limit leads (even none) and still have a next_cursor. Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time. next_cursor, passed back as cursor with the same filters and order, continues the list; next_cursor null means the list is complete. ids (up to 100 known ids, e.g. from webhooks or other lists) returns those records in one call, in the order given, with missing_ids for any not found. Outside card, the links returned are customer_url (a page the customer can open) and conversation.url (the thread in ShopBooster, for the shop's staff; texts containing it are refused with internal_link).

list_needs_attention

ChatGPT
The shop's stuck items in one list (leads and quotes that are waiting on the shop), and on the first page (no cursor) today {date, timezone, count, canceled, first (up to 5 of today's jobs: id, starts_at, customer, vehicle, services, techs), lines, more, summary}: today's appointments in shop time, read from the schedule. kind is awaiting_reply (the customer wrote after the shop's last message), quote_opened_no_followup (they opened a quote, not approved, declined or expired, and nobody has written to them since), approved_not_booked (quote approved, no appointment yet), invoice_overdue (balance due past the due date) or new_lead_not_contacted (a website form or Facebook/Instagram lead from the last 7 days, still in New Lead, that nobody from the shop has texted, called or emailed since it came in; listed once the shop's lead alert time has passed (Settings → Notifications → Lead alerts → Alert after; 5 minutes unless the shop changed it), the moment the team is alerted, unless older_than is given, "0m" for right away). Each item has a plain-language summary, value (quote total or balance due; for awaiting_reply their open quote, else null), waiting_since, customer, vehicle, proposal (status, total, customer_url, open_count, last_opened_at, deposit) or invoice (number, balance_due, due_date, customer_url), lead (id, stage; on new leads also type form or social, services and created_at), opted_out, last_inbound_at, last_outbound_at and conversation (source, id, path). New-lead items also have auto_text {status: scheduled, sending, sent, failed, skipped, cancelled_team_texted or off; at; reason, e.g. repeat_customer, opted_out, no_text_consent} and last_contact {type: text, call, social or email; at; by {kind: team or automation, user_id, name}} (null if the shop has not reached them), and scheduled_texts (texts queued to go out to them, with send_at). A lead whose automatic text went out but no person followed up stays listed with auto_text.status sent, after the leads nobody reached; include_auto_texted: false hides those. A new lead whose auto_text is scheduled has no thread until that text goes out. Texts to opted_out customers are refused. Items that already have an open Shop Bot suggestion (a suggested reply waiting in the thread, or an action waiting for a teammate's tap) are left out; include_suggested: true shows them with suggestion. Spam, duplicate and test leads and blocked threads are left out. Sort: recent (the default: the customer's latest message first, so today's customers lead even when old threads are still open), age (oldest waiting first) or value (biggest first); filter with kinds and older_than (e.g. "2h"). next_cursor, passed back as cursor with the same filters and order, continues the list; next_cursor null means the list is complete. Outside card, the links returned are customer_url (a page the customer can open) and conversation.url (the thread in ShopBooster, for the shop's staff; texts containing it are refused with internal_link).

list_pending_actions

ChatGPT
Bookings, reschedules, cancels and quote or invoice sends that a connection asked for and that wait for a teammate's tap in ShopBooster ("Bot actions need my tap"), newest first. Each has id (the same id as in the 202 answer), kind, status (pending, approved, edited = approved after a teammate changed it, dismissed, expired or outdated), summary, preview (the card the teammate sees), request (what was asked), acted_by, acted_at, changes and final (what was used when edited, e.g. another time), result_ids (appointment_id, proposal_id, invoice_id, message_ids) and outdated {code, message}. status defaults to pending; all lists every one. Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time. next_cursor, passed back as cursor with the same filters and order, continues the list; next_cursor null means the list is complete.

list_proposals

ChatGPT
Proposals (quotes), newest change first. Each has status (draft, sent, opened, approved, declined or expired), total and total_basis, packages with prices, open_count and last_opened_at, deposit {required, amount, paid}, sent_at, expires_at, customer, vehicle, lead_id and customer_url (the approval page to send the customer; null for drafts). conversation is the customer's latest thread (source, id, path); last_inbound_at and last_outbound_at say when the customer last wrote in and the shop last wrote out. Filters: status, booked (with status approved, booked: false lists approved quotes with no appointment yet), since (created at or after), tag (a tag the customer has), updated_since; sort: total (biggest first with order desc) instead of the newest change. card is a compact card of the first 5 (apps that show cards draw it; card.markdown is the same card as text). Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time. next_cursor, passed back as cursor with the same filters and order, continues the list; next_cursor null means the list is complete. ids (up to 100 known ids, e.g. from webhooks or other lists) returns those records in one call, in the order given, with missing_ids for any not found.

list_tags

ChatGPT
The shop's customer tags (its own labels, e.g. "VIP", "Fleet", "Wants dark tint"), A-Z, each with name, color and customers (how many customers have it). Tag names are unique in the shop, ignoring case, and work as the tag filter on customer searches and lists. Tags are internal and not shown to customers.

list_time_corrections

ChatGPT
Techs' requests to fix their recorded hours, with each edit and the paid time before and after. Read-only.

mark_conversation_read

ChatGPT
Marks a conversation read in the shop inbox (and clears its bell alerts). Nothing reaches the customer. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

move_lead_stage

ChatGPT
Moves a lead to a pipeline stage: New Lead, Not Quotable, Contacted, Proposal Sent, Followed Up, Approved, Booked or Lost. With Lost, reason and note record why (both optional). Logged on the customer's Activity tab as Shop Bot. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

reschedule_appointment

ChatGPT
Moves every part of a job to a new start (and to tech_id when given) after the same slot check as booking. ShopBooster sends the customer no "moved" message (customer_notified: false). approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

search_messages

ChatGPT
Texts and Facebook/Instagram messages matching a keyword search across every conversation in the shop, newest first. Every word of q has to appear in the message (case-insensitive, partial words match: "tint" finds "tinted"). Each match has snippet (the words wrapped in double asterisks), direction (in: from the customer, out: from the shop), sent_at, source, customer {id, name} and conversation {source, api_source, id, path, url}; the thread is read by its conversation.api_source and conversation.id. Blocked threads and threads deleted for good are left out; calls have no text. Filters: since and until (sent at or after since, before until), direction, source. Times come back in ISO 8601 with the shop's UTC offset; a time passed without an offset is read as shop time. next_cursor, passed back as cursor with the same filters and order, continues the list; next_cursor null means the list is complete.

send_invoice

ChatGPT
Sends the invoice link to the customer by email or text through the app's invoice delivery. It reaches the customer, so the texting refusals apply first: opted_out, blocked, quiet_hours (outside bot_send_hours), the daily text cap and texting_not_set_up. No card is charged. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

send_message

ChatGPT
Sends one reply in an existing thread as Shop Bot (SMS, or Facebook/Instagram inside Meta's 24-hour window). The send is refused, and nothing goes out, when: the customer is opted_out or blocked; the thread is archived (conversation_archived); it is outside the shop's bot_send_hours (quiet_hours, with retry_after); the shop's daily cap is used up (daily_limit_reached); the text links a page in the shop's app (internal_link, with the customer_url to use instead); it holds card or ID details (sensitive_info) or cussing (profanity); or it is the first text to a number the shop has not texted before and first_message_confirmed isn't true (first_message_needs_confirmation; first_message_limit_reached past the daily cap of first texts). While the shop has "Bots must check before sending" on (the default), a send without the check_id of a check of exactly this text is refused with check_required; check_not_found, check_text_mismatch, check_too_fast, check_used, check_expired, check_blocked and warnings_not_approved mean the check_id can't be used for this text. Other refusal codes: invalid_source, text_too_long, send_disabled, idempotency_conflict, texting_not_set_up, validation_error. The answer's delivery block says where the message stands now: queued or sent means not confirmed delivered, and delivery.delivered_confirmed turns true only once the carrier or Meta confirms delivery. notify_on asks for a webhook when the status changes.

send_proposal

ChatGPT
Sends a proposal to the customer with the shop's own text and email templates and moves the lead to Proposal Sent. channel is email, sms or both (default: the shop's template switches). It reaches the customer, so the texting refusals apply first: opted_out, blocked, quiet_hours (outside bot_send_hours), the daily text cap and texting_not_set_up. A switched-off channel is refused (channel_off). A proposal already sent is refused unless resend is true. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

snooze_conversation

ChatGPT
Hides a conversation from the inbox until an exact time (ISO time with offset, in the future, at most a year out), then it comes back. Nothing reaches the customer. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

suggest_reply

ChatGPT
Puts a suggested reply in a conversation as a draft. Nothing is sent to the customer: a teammate sees "Shop Bot suggested" in the thread and sends it (it goes out from them, through the app's normal checks), edits it first, or dismisses it. reply_style in the shop info shows how this shop writes and how it edits AI suggestions. One open suggestion per conversation per connection: a new one takes the place of the previous one, which is kept in the thread's history marked replaced (replaced_draft_ids); nothing is deleted. Suggestions expire after 7 days, and close as out of date (draft.outdated) when a teammate or the customer sends a new message first, or the customer's payments, invoices, proposals, appointments or lead change. Links to pages in the shop's app are refused (internal_link). Blocked or opted-out customers are refused. media_urls: SMS only, up to 5 photo links ShopBooster signed for this shop (from a conversation or a webhook); any other link is refused (media_not_allowed). source_event_id (the webhook event behind the suggestion, e.g. a message.received event id) is echoed in draft.* webhooks. approved_by and approved_at are optional here: a teammate approves by sending the suggestion in ShopBooster.

tag_customer

ChatGPT
Adds and/or removes the shop's tags on one customer (internal labels the team sees on the contact, the contacts list and Messages; not shown to customers). Tag names, any case: add ["VIP"], remove ["Lead"]. A name the shop doesn't have yet becomes a new tag (listed in created). Tags the customer already has (or doesn't have, for remove) are left alone. Returns the customer's tags after the change. approved_by and approved_at are required on a real call (validation_error without them) and are stored in the audit log; preview: true needs neither.

Capabilities

InteractiveWrites

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/shopbooster"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/shopbooster)