MCP App Store
by KeenEthics
Arketa icon

Arketa MCP Server for Claude and ChatGPT

By Arketa

Overview

Arketa is the business platform fitness and wellness studios run on: scheduling, bookings, memberships, payments and client management. This connector lets studio owners and staff ask questions about their own studio in plain English and get answers from their live Arketa data. Ask how revenue trended this quarter and what drove it, which classes fill fastest, who hasn't visited in 30 days, how many memberships canceled last month, or how your intro offer converts. Claude picks the right report, pulls the data, and explains what it found. What it can read: 30+ report datasets (sales, bookings, attendance, subscriptions, intro offers, retention, gift cards), your live class schedule and capacity, your services and pricing catalog, studio locations and instructors, and client profiles by name or email. It is strictly read-only. It cannot book, cancel, charge, message clients or change anything. Each person connects with their own Arketa login and sees only the studio and the reports their staff role allows; nothing is shared between studios. Staff pay rates, payroll and payment card details are never exposed. Requires an Arketa studio account with permission to view reports.

How to connect

  1. Claude: Open the Arketa listing in the Claude connectors directory and choose Connect. Claude listing Arketa exposes 18 tools here.
  2. ChatGPT: Open the Arketa listing in the ChatGPT app directory and choose Connect. ChatGPT listing Arketa exposes 25 tools here.

Tools

client_overview

ChatGPT
Everything about one client in one call, read live: profile and contact reachability (canEmail/canText), active memberships with their live standing (status, next billing, cancel/pause), credits left across packs and memberships, last visit and next booking, outstanding invoices, and spend. Use it for 'tell me about Jane', 'does she have credits left', 'is her membership paid up'. Use search_clients to find the clientId. Outstanding invoices and spend are left out (and named in omittedForRole) when the signed-in role cannot see them. Money is in the studio currency's main unit; never divide it.

clients_at_risk

ChatGPT
Who might not renew: failed membership payments, pending cancellations, contracts ending, lapsed regulars, unused packs about to expire, intro offers not converting, and clients the churn score flags. Returns up to 25 people ordered by their most urgent reason, each with reason codes, last visit, membership, canEmail and canText, plus totals and the plain-English rule per reason. Filter with reasons (e.g. ["past_due"] for failed payments, ["lapsed_regular"] for regulars who stopped coming). Use get_clients only if you need more contact detail than this. Scoped to the authenticated studio.

describe_report

ChatGPT
Describe a single report dataset: the filterable fields (with their type and valid operators), the date field that a startDate/endDate range is applied to, and the numeric (aggregate) and segment fields available for grouping. Use this before run_report to build valid filters.

find_payment

ChatGPT
Look up a specific payment/charge by ONE identifier: a Stripe invoice id (in_...), an Arketa order id, or a payment id. Searches the studio's full payment history (all statuses) and returns the matching transaction rows PLUS the purchase the charge belongs to — including when that purchase lives on a linked FAMILY MEMBER's profile while billing this payer (a recurring charge with no matching purchase on the payer's own profile is usually this, not an orphaned charge). Human-facing invoice NUMBERS like ABCD1234-0017 are not stored — provide the in_... invoice id instead, or find the charge via run_report (salesAll) by date and amount. Amounts are in the studio currency's main unit. Scoped to the authenticated studio.

get_class

ChatGPT
Get a single scheduled class by id (use get_schedule to find ids): time, instructor, location, capacity, spots remaining, waitlist length, description, booking flags and pricing option names. Scoped to the authenticated studio.

get_client

ChatGPT
Get a single client's profile by id (use search_clients to find the id), for one client; for several ids use get_clients. Returns the client's key profile fields, including their current tags as { id, name } (the ids tag_clients takes). Scoped to the authenticated studio.

get_client_credit_history

ChatGPT
How one of a client's memberships or packs got and spent its credits, newest first: each cycle's credits granted (granted, with the allowance and the cycle it opened), each class booked (booked, -1, named), each cancellation that returned a credit (returned, +1), and any staff adjustment (adjusted), with the balance after each. A canceled booking with no returned entry after it means the credit was not put back. purchaseId comes from get_client_purchases; clientId from search_clients/get_client.

get_client_forms

ChatGPT
Get the forms a single client has filled out (waivers, intake forms, questionnaires) with their submitted answers and submission dates. Use search_clients/get_client to find the clientId. Returns one entry per submission across all of the studio's forms. For the blank form catalog (what forms exist) use list_forms; for every client's responses use list_form_responses.

get_client_purchases

ChatGPT
Get one client's purchases (packs, memberships/subscriptions, bundles, drop-ins) with offering name/type, price, credits remaining, subscription status and expiry. Use search_clients/get_client to find the clientId. Defaults to the full completed-purchase history (most recent first); pass activeOnly=true for only currently-usable purchases. price is in the studio currency's main unit (e.g. dollars, exactly as stored — a $150/month plan is price 150; never divide it). Also returns familyMemberPurchases: purchases on linked family members' profiles, which do NOT appear in this client's own list even though their recurring charges usually bill this client (the family-account owner) — check these before calling a charge orphaned. This is per-client history; for studio-wide sales totals use run_report (salesAll / salesSuccessful).

get_client_reservations

ChatGPT
Get one client's reservations (booking history) — per booking: class name, start time, instructor, location, reservation type, check-in/attendance state, cancellation/late-cancel flags, waitlist position and whether it was paid. Use search_clients/get_client to find the clientId. This is per-client history; for studio-wide attendance use run_report (reservationsAll = bookings, classCheckIns = attendance). Most recent first; pass startDate/endDate to bound the range.

get_clients

ChatGPT
Profiles for up to 100 client ids in one call: use this after a report returns client ids, never get_client once per id. Returns a compact summary per client (name, email, phone, marketingOptIn, tags as { id, name }, created / last visit) plus smsReachable and emailReachable, which say whether a text or an email can reach that client. Ids that are deleted, removed, anonymous, a walk-in or unknown come back in missingIds. If the result is too large, droppedIds names the ids left out: call get_clients again with exactly those. Scoped to the authenticated studio.

get_family_accounts

ChatGPT
Get the family-account links around one client, in BOTH directions: familyMembers are the people linked under this client's account (this client is the account owner — a family member's subscription lives on the FAMILY MEMBER's profile, while the recurring charge usually bills the OWNER's payment method), and memberOf are the accounts that list this client as a family member (that owner is usually who pays for this client). ALWAYS check this when a client has a recurring charge with no matching purchase on their own profile — it is usually a family member's subscription, not an orphaned charge. Use search_clients/get_client to find the clientId. Scoped to the authenticated studio.

get_schedule

ChatGPT
Get the studio's class schedule (the live calendar, including FUTURE classes) for a date range, with instructor, location, capacity, booked count and spots remaining per class. Dates are interpreted in the studio's timezone. For historical revenue/attendance analysis use run_report instead — this tool is the source of truth for what is scheduled. The name of a class here matches the report filter class_name.

get_studio_context

ChatGPT
Get the studio's operating context: identity (name, timezone, currency), locations, instructors/staff, service categories, and client tags. Call this early — the instructor ids returned here are the ids used in report filters (instructor_id / instructor_ids_array) and in the schedule's instructorIds (host ids), the location ids match the report filter location_id, and the tag ids are what tag_clients takes (a tag the studio has not created is not listed and cannot be added). Scoped to the authenticated studio.

list_form_responses

ChatGPT
List form submissions (responses) for the studio — by default across all forms, or pass formId (from list_forms) to scope to one form. Each response has the form name, respondent (name/email), submission date and the submitted answers. Most recent first; results are capped to fit the context window, so pass formId and a small limit to narrow. For one client's submissions use get_client_forms; for the blank form definitions use list_forms.

list_forms

ChatGPT
List the studio's configured forms — waivers, intake forms and questionnaires — with each form's id, name, enabled state, question schema (label/type/required) and what it applies to (services, categories, offerings). This is the blank form catalog (configuration), not submitted answers. Use the returned formId with list_form_responses to read responses for one form. Excludes deleted forms by default.

list_offerings

ChatGPT
List the studio's configured pricing options — class packs, subscriptions/memberships, bundles and challenges — with price, credits, expiry and billing cadence. price is in the studio currency's main unit (e.g. dollars, exactly as configured — a $150/month plan is price 150; never divide it). This is the configured catalog of what's on sale (it joins to salesByPricingOption report rows by name), not a recommendation engine and not sales history — for what actually sold, use run_report.

list_reports

ChatGPT
Fallback catalog of report datasets for this studio, grouped by business domain (sales, reservations, memberships, intro offers, clients/retention, pricing, gift cards). The full catalog is already embedded in your system instructions — do NOT call this unless that catalog appears missing or corrupted. When you do need it, match the user's question to a report's whenToUse / exampleQuestions / synonyms, then call describe_report for filters, then run_report for data.

list_services

ChatGPT
List the studio's configured services — the class/appointment/event types that classes are scheduled from — with duration, price, capacity, category and assigned locations/instructors. These name values are the valid class_name filter values for reports. This is configuration, not the calendar: for what is actually scheduled (dates, spots remaining) use get_schedule.

run_report

ChatGPT
Run a report for this studio and return aggregate totals plus a sample of raw data rows. Provide a date range (startDate/endDate, applied to the report's main date field) and/or explicit filters. For any total / average / rate / trend / top-N question, read the aggregates (they reflect the FULL result) instead of pulling rows to compute it yourself. When you do need rows, always pass fields to project to just the columns you need and keep limit small — rows are capped to fit the context window, so requesting everything will not return everything. Always scoped to the authenticated studio. Use list_reports + describe_report first to choose a report and valid filters.

sales_summary

ChatGPT
Sales for a period broken down by category, pricing option and client source, plus refunds (with the offerings refunded most) and tips, each against a comparison period. Use it for 'sales by category vs last month', 'which pricing options sold most', 'how much did we refund'. Rows are the top 25 by amount (total covers all). Amounts are in the studio currency's main unit. The source breakdown is revenue by the client's acquisition source, windowed by when the client was created, not when they paid. Breakdowns the user's role cannot read are listed in omittedForRole.

schedule_attention

ChatGPT
Under-filled classes, classes with waitlists, classes with no teacher, and open substitution requests for the next few days, read live from the schedule. Use it for 'what needs attention in the next 3 days', 'any classes with no teacher', 'which classes have waitlists'. Canceled and on-demand classes and availability time blocks are excluded; 1:1 appointments never count as under-filled. Counts are complete; each list shows the first 25 by start time.

search_clients

ChatGPT
Find clients in this studio. Provide email for an exact lookup, or query for a fuzzy search across name / email / phone. Returns a compact summary per client (including their id, which you pass to get_client, and their current tags as { id, name }). Name/phone searches use a search index that can be incomplete (e.g. migrated studios); on 0 matches the tool automatically retries with a direct scan of the client list — but a 0-match result STILL does not prove the client doesn't exist. Never tell the user a client is not in their records from a failed name search alone: ask for the client's email and retry with email. Scoped to the authenticated studio.

studio_brief

ChatGPT
One-call weekly brief of what needs the owner's attention, ranked by estimated dollars at stake: payments failing, cancellations and contracts ending, lapsed regulars, unused packs, intro offers going cold, classes losing attendance. Returns counts and the rule behind each finding, never names. Use it for "what should I focus on", "Monday brief" or "is anything slipping". Each finding's dollar figure is an estimate from its own rule; they overlap, so never add them together. For the people behind a finding, call clients_at_risk with the finding's reason. Scoped to the authenticated studio.

studio_kpis

ChatGPT
One-call scorecard for a period: gross and net revenue, bookings, attendance, new leads, new clients, new subscriptions, intro offer conversion and MRR by month, each with the previous value, change and the report it came from. Use it for 'how did last month go', 'compare October to September', 'how many new clients this month'. Revenue is in the studio currency's main unit. MRR is by month (not by day) and may not include v2 billing-schedule memberships: read coverageNotes. Metrics the user's role cannot read are listed in omittedForRole.

describe_report

Claude

find_payment

Claude

get_class

Claude

get_client

Claude

get_client_forms

Claude

get_client_purchases

Claude

get_client_reservations

Claude

get_clients

Claude

get_family_accounts

Claude

get_schedule

Claude

get_studio_context

Claude

list_form_responses

Claude

list_forms

Claude

list_offerings

Claude

list_reports

Claude

list_services

Claude

run_report

Claude

search_clients

Claude

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/arketa"><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/arketa)