MCP App Store
by KeenEthics
Horex AI icon

Horex AI MCP Server for ChatGPT

By Horex

Overview

Horex AI helps restaurant owners and managers review operations across their restaurant group using the HoreX records their account is permitted to access. Compare sales, profit, food cost, procurement and stock across branches and periods. Find Finance and Catering documents, inspect POS shifts and recorded attendance, and review Central Kitchen evidence. Explore customer cohorts, RFM, geography, advertising, broadcasts and automation results. Investigate Inbox issues through saved summaries and selected bounded excerpts. Receive native ChatGPT answers with links to the exact report or record in HoreX. Coverage gaps, incomplete periods and missing evidence remain explicit. Prepare Team task proposals and campaign settings in the conversation, then review and approve them in HoreX. Campaign preparation covers new Meta campaigns and ordinary Google Search campaigns, plus supported launch, pause and budget changes to existing campaigns. Geo Marketing uses its separate workspace. Campaign settings are saved for native approval before provider execution. Available actions depend on source permissions and connected advertising accounts. The plugin does not directly edit invoices, stock, recipes or payments, send broadcasts or make purchases. An existing HoreX account, an enabled organization and the ChatGPT integration permission are required. Existing branch, source and action permissions apply. Access is currently piloted with selected organizations.

How to connect

  1. ChatGPT: Open the Horex AI listing in the ChatGPT app directory and choose Connect. ChatGPT listing Horex AI exposes 62 tools here.

Tools

find_google_ads_campaigns

ChatGPT
Find existing ordinary Google Ads campaigns by saved name and status. Returns typed google_ads: campaign_ref values for review; excludes Geo Marketing campaigns. Resolve ambiguous names before preparing a launch. Costs 1 AI Credit.

find_meta_campaigns

ChatGPT
Find existing Meta campaigns by saved name and status. Resolve ambiguous names before preparing a launch. Costs 1 AI Credit.

find_restaurant_entities

ChatGPT
Resolve product, supplier, brand (concept), branch or warehouse names to visible IDs and units. Ambiguous candidates require clarification. Costs 1 AI Credit.

get_advertising_status

ChatGPT
Read cached warehouse freshness and sync status without refreshing providers or changing stale jobs. No credentials or raw provider errors. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

get_analysis_catalog

ChatGPT
Free: returns available analytical datasets, measure codes, dimensions, filters, units and coverage. Read before composing query_restaurant_data.

get_catering_document

ChatGPT
Reads one native Catering invoice or quotation with line quantities, units, prices, VAT, payments and available cost evidence. Use the Catering ID and type returned by list_catering_documents. Finance IDs belong to another source. Contacts, bank details, internal notes and payment links are excluded. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_central_kitchen_day

ChatGPT
Read native Central Kitchen orders, delivery differences and, only with planning/cooking access to the supplying kitchen, production batches, output adjustments and attention. Explicit Dubai date. Outstanding older work is separate from the selected delivery date. Preserve units and coverage.

get_central_kitchen_delivery

ChatGPT
Read an exact permitted Central Kitchen delivery and packed/received differences. Native receiving-branch and kitchen rules apply. This tool does not receive, dispatch or resolve a difference.

get_central_kitchen_item

ChatGPT
Inspect one native Central Kitchen item: branch demand, confirmed allocations, stock availability, production batches, actual output reconciliation, recipe and current estimated cost. Requires plan/cook access to the supplying kitchen. Missing costs are unknown; below-plan output is not proof of physical waste. Never conduct, ship or adjust stock.

get_central_kitchen_order

ChatGPT
Read an exact permitted Central Kitchen branch order and requested, confirmed, shipped and cancelled quantities, with its native source link.

get_central_kitchen_workspace

ChatGPT
Resolve visible native Central Kitchens and connected branches. A branch user cannot see the kitchen internal production or cost. Choose an exact kitchen for follow-ups.

get_daily_review

ChatGPT
Free: opens the saved daily report referenced by a daily_review.ready event. The original scheduled report was charged once; reopening is free. This is a dated snapshot, not a new live read. Current report and branch access are rechecked. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_finance_invoice

ChatGPT
Reads one permitted finance invoice and its line amounts. Excludes contacts, bank data and payment links. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_food_cost_drivers

ChatGPT
Explain food cost using historical sold quantities, verified sold-line costs and ingredient price/mix reconstruction from the same internal Copilot service. Returns baseline/current coverage, ranked dish and ingredient drivers and source links. Reconstruction uses current recipe composition and must not be called exact historical consumption. Omitted baseline means the preceding equal-length period. Costs 1 AI Credit.

get_food_cost_variants

ChatGPT
Compare posted-cost basis at period cutoffs with ingredient evidence. Supplementary Batch Breakdown measured-yield costs and estimated inventory gap allocation preserve primary food cost. Includes exact document links, formulas and coverage. Inventory estimate is not dish-level actual food cost or a proven cause; never add batch and inventory deltas. Maximum 92 days and 50 recipes; narrow source-heavy requests. Costs 1 AI Credit.

get_google_ads_campaign_draft

ChatGPT
Read your saved campaign draft and native approval result. Can read settings, blockers and status; cannot publish, resend or reconcile a provider write. confirmed means settings match, not delivery. Each successful call costs 1 AI Credit.

get_google_ads_campaign_operation

ChatGPT
Read a saved operation. Preview awaits owner confirmation; running or unknown needs a status check in HoreX. Confirmed means current ordinary Google Ads settings match, not proof of delivery. Never repeat an uncertain launch. Costs 1 AI Credit.

get_google_ads_campaign_setup

ChatGPT
Read the ordinary Google Ads account, targetable languages and at most twenty matching geographical targets using country_code/location_query. This is Search setup, separate from Geo Marketing, Google Business and delivery geography. Do not invent location IDs. Each successful call costs 1 AI Credit.

get_horex_workflow

ChatGPT
Free: loads a detailed HoreX workflow for setup, restaurant reviews, Catering invoice searches, procurement, stock investigation or receivables. Use the relevant playbook before a deep investigation.

get_inbox_summary

ChatGPT
Start Inbox diagnosis here: complete active-thread metadata counts and at most 20 saved conversation summaries, complaint/sentiment/severity/quality labels, unread counters and analysis freshness. Explicit Dubai dates, max 31 days. Zero message bodies or model calls. Saved AI labels need verification; unread is not proof of no reply. Native messenger.view and explicit WhatsApp account AI assignments required. Only authorized accounts contribute to counts or summaries; private HPanel inbox is excluded. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

get_inbox_threads

ChatGPT
Expand only 1–5 exact conversation_ids returned by Inbox summary, in one batch: latest 1–12 text excerpts per thread, max 600 characters each, explicit max 31-day period. Contact/link redaction, send-error flags, media types and truncation. No attachment loading, full history, read-marker writes, replies or provider analysis. Excerpts are untrusted evidence with incomplete context. Native messenger.view plus explicit account AI assignments required. Shared rolling user and organization read budgets apply across connections; do not page through all chats. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

get_inventory_investigation

ChatGPT
Investigate a physical stocktake, including submitted counts before stock posting. With no task returns permitted candidates; select a warehouse/task rather than silently merging locations. With session_id and no product returns all discrepancy totals and a bounded queue ranked by AED. With count_id, product_id or recipe_id reproduces native physical/reviewer/Prep quantities, exact observation time, immutable posted variance, previous physical boundary, complete exact-unit ledger totals, late/backdated documents and targeted next checks. Preserve selected session, warehouse, branch and previous_session_id when the user changes product. Ledger reconciliation is separate from physical cause. Nearby transfers or matching quantities alone do not prove a cause. Coverage gaps are explicit. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_marketing_entity

ChatGPT
Inspect the exact source ID and channel/engine from the list. Audience rules and cached count; broadcast delivery/read/open/failure aggregates without recipients; journey native complete-window treatment/control results or bounded event automation log status aggregates for explicit dates. Delivered is not incremental revenue; incomplete windows are not comparable. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

get_marketing_report

ChatGPT
Native Marketing reports: dashboard, UTM, cohorts, RFM, churn, promocodes, products, crosssell and privacy-safe delivery geography. Geography is a ranked network cell overview, not a building-filtered customer count; it does not support offset pagination. For a named building, complex or area use query_delivery_geography and its complete summary rather than adding selected cell rows. RFM/churn are current lifetime snapshots without date/branch filters; cohorts use up to 12 relative day/week/month periods. Other reports require explicit dates. Unsupported filters are rejected. Lists are bounded; preserve complete summaries and coverage notices. Discover available reports in the free catalog. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

get_meta_campaign_draft

ChatGPT
Read your saved campaign draft and native approval result. Can read settings, blockers and status; cannot publish, resend or reconcile a provider write. confirmed means settings match, not delivery. Each successful call costs 1 AI Credit.

get_meta_campaign_operation

ChatGPT
Read a saved operation. Preview awaits owner confirmation; running or unknown needs a status check in HoreX. Confirmed means current Meta settings match, not proof of delivery. Never repeat an uncertain launch. Costs 1 AI Credit.

get_meta_campaign_setup

ChatGPT
Read the connected Meta account, permitted pages, existing images, custom audiences and purchase pixels in one bounded lookup. At most twenty choices per source; has_more means incomplete. Do not invent IDs or fetch all account assets. Each successful call costs 1 AI Credit.

get_pnl

ChatGPT
Returns the canonical HoreX P&L summary, COGS method and report calculation time. Ordinary reads may reuse a snapshot for up to 120 seconds. Set refresh true when the user asks to recheck or has just changed payroll or other source records; it forces one native recalculation. data.freshness.calculated_at is the report time; source.retrieved_at is the response time. If freshness is absent, calculation time is unknown. Payroll includes only approved/paid records within the returned scope; zero alone does not prove missing salaries. Use explicit Dubai dates, at most 92 days. Costs 1 AI Credit, including an explicit refresh. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_pos_shift

ChatGPT
Read one visible POS cash shift by the ID returned by get_pos_shifts. Returns its lifecycle, cashier, saved close counts, expected/actual cash, variance, native tender breakdown and bounded cash operations. Unclosed or uncounted cash variance remains unknown. Does not close, reopen or alter a shift. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_pos_shifts

ChatGPT
Check whether cash-register shifts are closed. Reads native POS shifts including shifts without orders, exact lifecycle counts across all matching visible shifts, opening/closing times, terminal, Z-report number and saved cash variance. Use status all with explicit Dubai from/to dates to check a day; date_basis active_during includes overnight shifts active in that period, opened matches Shift Reports opening dates. To find all currently open shifts including older unclosed shifts, use status open without dates. Maximum 92 days and 100 displayed shifts; summary precedes row limits. Closed orders do not prove a closed shift. These are POS cash shifts, not HR attendance. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_profile

ChatGPT
Returns the connected account and organization. Connection checks are free.

get_profit_review

ChatGPT
Compare canonical P&L periods and investigate profit changes. Returns expense breakdowns, a reconciled profit bridge, branch sales comparisons and ranked investigation candidates. Candidates are observed differences, not guaranteed savings or causal findings. Omitted baseline means the preceding equal-length period. Both periods cost 1 AI Credit together. Read profit-improvement guidance for goals and opportunity discovery.

get_receivables

ChatGPT
Returns an exact summary over all matching visible unpaid or overdue Finance invoices plus up to 100 recent records. Omit dates for all-date debt; supplied dates filter invoice date. Owner and branch visibility apply. Rows are not a full aging report when truncated. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_recipe_economics

ChatGPT
Inspect a named recipe or prep with canonical recursive current cost, selling price, contribution, missing cost evidence and ingredient/prep components. Search is required to keep the read focused. Current planning cost is distinct from historical sold-line COGS. Costs 1 AI Credit.

get_sales_summary

ChatGPT
Returns sales from the canonical P&L source, including channel and branch breakdown. Use explicit Dubai dates, at most 92 days. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_staff_attendance

ChatGPT
Answer who arrived late using canonical HR schedules and recorded check-ins. Required explicit Dubai from/to dates, max 92 days. Use status late for late arrivals; missing means no recorded check-in after planned start, not proven absence or lateness. Unknown schedule means lateness cannot be established. Includes full scope counts before status filter and row limit, distinct late people and bounded employee/date rows with planned time, check-in and minutes late. A late employee can still have an open attendance record. branch_id means employee home branch, not physical check-in location. Approved leave and days off do not count as late. Stored schedule records only; unscheduled staff/events are not an exhaustive roster. Today is partial. No payroll or disciplinary decisions. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_stock_balances

ChatGPT
Returns a page of physical stock and its units. Do not combine incompatible units. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

get_team_task

ChatGPT
Track one permitted native Team task, its assignees, current status, discussion and activity. Task completion is separate from measured business improvement; compare fresh operational facts against the recorded baseline and acceptance criteria.

get_team_workspace

ChatGPT
Search the minimal eligible Team directory and accessible projects before choosing an assignee. Existing private project, participant and observer permissions apply. Never invent a person or expand project membership.

list_branches

ChatGPT
Lists accessible HoreX branches. Costs 1 AI Credit per successful request. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

list_catering_documents

ChatGPT
Searches native Catering invoices or quotations by invoice/document number, customer or company. Owner, branch and linked-order access apply. With no dates searches all dates; date filters use event_date by default, or explicitly created_at. Up to 100 rows per page, matching count and has_more included. Finance invoices are a separate source. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

list_customer_profiles

ChatGPT
Bounded named customer frequency, value, recency and lifecycle lists, including lapsed customers. No contacts or raw customer IDs. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

list_finance_invoices

ChatGPT
Searches Finance invoices only, not Catering invoices. For unspecified customer invoice searches check both Finance and list_catering_documents. Returns up to 100 recent matching records, not an all-time debt balance. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

list_marketing_entities

ChatGPT
Search saved audiences, real broadcasts (explicit whatsapp/push/locus_push channel) or automations (journey/automation/loyalty_journey engine). Complete matching counts and deterministic pagination. This does not send, activate or edit anything. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

list_team_tasks

ChatGPT
Read native Team tasks with full matching counts, bounded rows and explicit pagination. Private task and project audiences apply. Reuse existing work rather than propose duplicate tasks.

open_restaurant_workspace

ChatGPT
Free: returns the current permitted HoreX workspace, available report choices and default Dubai dates as facts for a native reply in the current AI application. Widget view is paused; use normal conversation text and tables. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

prepare_google_ads_campaign_launch

ChatGPT
Prepare and SAVE an exact existing campaign change in HoreX, including requested budget and explicit paused child selections. Resolve missing selections in chat using review first. Returns the saved operation and its final approval URL; no setup wizard or re-entry is needed. Local preview only: no ads are changed or activated. Use a UUID request_key and reuse it only for an identical retry. The user approves the saved settings in HoreX.

prepare_meta_campaign_launch

ChatGPT
Prepare and SAVE an exact existing campaign change in HoreX, including requested budget and explicit paused child selections. Resolve missing selections in chat using review first. Returns the saved operation and its final approval URL; no setup wizard or re-entry is needed. Local preview only: no ads are changed or activated. Use a UUID request_key and reuse it only for an identical retry. The user approves the saved settings in HoreX.

prepare_new_google_ads_campaign

ChatGPT
Save or revise a COMPLETE new ordinary Google Search campaign in HoreX. Include geo and language targets, exact/phrase keywords, explicit negatives, manual CPC, Search ad groups, headlines, descriptions, final URLs, budget and the returned account currency. starts_on/ends_on are calendar dates in the account timezone; ends_on is inclusive. Resolve missing facts with the user in chat; never invent assets, billing access or tracking. New drafts use draft_id=null/version=null. Revisions use returned draft_id/version. Generate one UUID request_key per preparation and reuse it only for an identical retry. Saves a local, actor-bound, versioned draft and readiness review; does not create external objects or spend. Summarize the full saved settings and give approval_url ONLY for final human approval in HoreX; do not send the user through setup again. Read the result after approval. Never claim ads launched from draft state. Each successful call costs 1 AI Credit.

prepare_new_meta_campaign

ChatGPT
Save or revise a COMPLETE new Meta traffic/sales campaign in HoreX. Include geography, permitted custom audiences, an existing Meta image, allowed page and purchase pixel where applicable, full ad copy, links, schedule, budget and the returned account currency. Meta starts_at/ends_at use explicit timestamp offsets. Resolve missing facts with the user in chat; never invent assets, billing access or tracking. New drafts use draft_id=null/version=null. Revisions use returned draft_id/version. Generate one UUID request_key per preparation and reuse it only for an identical retry. Saves a local, actor-bound, versioned draft and readiness review; does not create external objects or spend. Summarize the full saved settings and give approval_url ONLY for final human approval in HoreX; do not send the user through setup again. Read the result after approval. Never claim ads launched from draft state. Each successful call costs 1 AI Credit.

prepare_team_task

ChatGPT
Prepare one immutable task proposal for the user to review in HoreX. Resolve assignees first; include concrete work, due date, acceptance criteria and observed baseline. Returns a review link, never creates or assigns a task. Only the proposing user can explicitly confirm in the authenticated HoreX page; fresh access is checked again. Do not claim assignment before committed status.

query_ad_performance

ChatGPT
Bounded cached campaign/adset/ad performance with complete matching totals separated by currency. One hierarchy level per call; never add campaign and ad spend together. No sync or provider calls. Platform conversions differ from HoreX recognized sales. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

query_ads_acquisition_metrics

ChatGPT
Compare Meta/Google warehouse spend, platform conversions, UTM orders/revenue and new-to-brand versus new-to-company CAC. Meta includes Facebook and Instagram. Unknown attribution/spend is not zero. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

query_blended_acquisition_metrics

ChatGPT
Blended CAC from confirmed marketing costs and new customers irrespective of UTM. Choose new_to_organization or new_to_brand; preserve unallocated cost and possible invoice/platform overlap. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

query_customer_metrics

ChatGPT
Customer totals, new/returning/repeat rates, frequency, average check and identification coverage. Returning means earlier visible purchase; repeat means two purchases in the selected period. Distinct customers cannot be summed across groups. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

query_customer_order_composition

ChatGPT
Customer purchase mix or anchored co-purchases. For bought together use analysis_mode co_purchase, population all_orders, dimension item and item_name. Modifier revenue is not additive. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

query_customer_profitability

ChatGPT
Historical frozen order-line COGS and covered gross margin by customer/item/category. Preserve coverage; current recipe cost is not historical COGS. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

query_delivery_geography

ChatGPT
Find recognized deliveries to a named building, residential complex or area from saved delivery location text. Required explicit Dubai dates and 1–5 specific place_queries (literal alternatives, minimum 3 characters); optional area_query intersects the match. One set-based read returns complete distinct identified customer count, orders, revenue, repeat customers, identification/search coverage, scoped delivery share and up to 50 approximate cells. Branch access applies. No raw addresses or contacts. Text matches do not prove all towers/residents; do not invent aliases or expand coordinates. Unknown identity is excluded from customer counts. Repeat is at least two matching deliveries within the period, not lifetime retention. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

query_geo_marketing

ChatGPT
Read stored Google Geo Marketing campaigns and restaurant Business Profile locations, not delivery destinations, customer buildings or residential complexes. For customers from a building/area use query_delivery_geography. View locations accepts no dates; campaigns/campaign require explicit dates. Scoped to accessible branches. No Google refresh, publish, budget changes or live Business Profile performance API. Missing cached metrics are unavailable, not zero. Dated marketing reads cover at most 92 days per period; detailed lists have at most 50 rows. Respect source-specific date modes. Each successful GPT business read costs 1 AI Credit; native customer/CAC tools can compare two explicit periods in one read.

query_restaurant_data

ChatGPT
Aggregate permitted restaurant facts and compare two periods in one request. Discover codes with get_analysis_catalog. Explicit Dubai dates, max 92 days per period, max 100 rows per period. Totals precede row limits; preserve truncation and units. Current stock has no date range. Costs 1 AI Credit. For Open in HoreX use the returned source_links verbatim; they preserve report filters or the exact record. Do not substitute a generic page URL.

review_google_ads_campaign_launch

ChatGPT
Use the google_ads: campaign_ref returned by find_google_ads_campaigns. Bare IDs and Geo references are rejected. Read current ordinary Google Ads account, campaign, ads, schedules and budget and return the HoreX approval link. Does not launch or change ads. Owner confirms exact settings in HoreX. No external plugin needed. Existing Google Ads campaigns only; excludes Geo Marketing. Up to ten ad/asset groups and fifty ads. Shared/total budget edits blocked; Performance Max has asset groups. Costs 1 AI Credit.

review_meta_campaign_launch

ChatGPT
Read current Meta account, campaign, ads, schedules and budget and return the HoreX approval link. Does not launch or change ads. Owner confirms exact settings in HoreX. No external plugin needed. Existing AED campaigns only, up to ten ad sets and fifty ads. Costs 1 AI Credit.

Capabilities

Writes

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/horex-ai"><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/horex-ai)