Connect
https://mcp.tacet.ai/mcpCustom connectors work on every Claude plan. On a Team or Enterprise plan an Owner adds the connector once and each member then connects with their own Tacet account. Claude Code works too, with the command below. Questions to start with are on the Tacet in Claude page of your portal.
In Claude.ai
- Open Settings, then Connectors, then Add custom connector.
- Name it Tacet and paste the URL above.
- Choose Connect, sign in with your Tacet account and pick your store.
In Claude Code
claude mcp add --transport http tacet https://mcp.tacet.ai/mcp
Then run /mcp inside Claude Code and authenticate.
Who it works for
A Shopify store on Klaviyo that has installed Tacet from the Shopify App Store. Sign-in uses the same account as the Tacet portal. If your store is not on Tacet yet, follow the setup steps first; a new store's first numbers usually arrive about 20 minutes after installing.
What Claude can ask
34 read-only tools. Every answer carries the scoring run's date and what the numbers rest on.
get_value_overview | The store's value overview from the latest scoring run: total customers, total Forward CLV (what the customers will be worth over the next 12 months, after product cost and expected returns, valued in today's dollars), total gross CLV, the ratio between them, value tiers, concentration, archetypes and behavioral profiles. |
get_what_changed | What changed between scoring runs: metric deltas with a population-versus-per-customer decomposition, plus ranked insight items. |
get_trajectory | Time series across scoring runs: total Forward CLV, gross CLV, customers, average return propensity, concentration and repeat buyers, one point per month. |
get_forecast | The forward line: what the customers scored on the latest run are expected to buy and contribute, by week for thirteen weeks and by month for twelve, each with a p10 and p90 band, plus an acquisition layer at a stated weekly first-order rate priced on the payback curve. |
get_growth_levers | Ranked growth levers: concrete opportunities with an estimated dollar impact and a target population, from the latest scoring run. |
get_plays | The ranked growth levers with the plays run on them: a play is a lever's audience sent to the merchant's mailing tool with a share held back, measured weekly from the orders that follow. |
get_cohorts | Acquisition cohorts on TRUE first-purchase dates: every customer grouped by the month of their first ever order, with the share of each group that ordered again N months later, measured across the store's whole order history. |
get_product_scopes | Every product type, collection and tag that the customer and cohort filters will match, each with its product count, read from the catalog on every call. |
list_customers | Paginated customer list with per-customer scores: Forward CLV, gross CLV, the chance each customer is still active, predicted and past return rates, RFM, margin, subscription candidacy and reorder timing. |
find_customer | Find ONE customer by exact email address, numeric customer id, or Shopify order number prefixed with its hash sign. |
get_customer | Full score profile for one customer by customer id: Forward CLV, gross CLV, the chance they are still active, expected purchases, RFM, return propensity, promo reliance, margin, tier and percentile bands, plus realized revenue, contribution and orders to date. |
get_product_value | Products ranked by the average Forward CLV of the customers who buy them, with how those buyers behave afterwards (return rate, promo reliance, one-timer share) plus units, price and COGS. |
get_segment_patterns | One of three segment analyses. |
get_returns_concentration | How unevenly returned DOLLARS land across the customer base, measured on refunds already issued: history, not prediction, and distinct from return propensity, which predicts. |
get_returns_distribution | How return rates are SPREAD across the base, and return rate by Forward-CLV decile (decile 1 least valuable, 10 most). |
get_data_status | The state of this store's data: latest scoring run id and observation date, whether it is fresh, integration status (Shopify, Klaviyo, returns) with last sync times and errors, and the scoring settings behind the numbers (the margin the store gave us, if it has, beside Tacet's default margin, which prices nothing once the store has stated one; discount rate, return retention rate, fee and fulfillment rates). |
get_payback | Payback curves: cumulative contribution per acquired customer at months 0 to 12 after their first order, for the whole base, for each gateway product (customers whose FIRST order contained it) and for recent acquisition cohorts. |
get_reorder_timing | The pull-forward population, sized: per reorder band (due now, overdue, at risk, on schedule) how many customers and the Forward CLV they hold, with median days since last order, the share on their own measured cadence and mean still-active probability, plus how many customers the measurement covers at all. |
get_sensitivities | What moves the number: one-unit sensitivities of total Forward CLV to six levers (one point of one-timer conversion, one percent more purchases, promotions one point shallower, one point of margin, one point less return propensity, one hundred more acquired customers), each with the change in Forward CLV, its share of the total, the customers it touches and a sentence saying what was held fixed. |
get_data_gaps | What Tacet is missing from this store, ranked by how much it moves the numbers: product costs not on file (and the share of line items with no cost of their own, with what priced them: the store's own costed products, the margin the store gave us, or a default), payment and fulfillment costs not set, returns read from Shopify refunds only (so exchanges are invisible and every returned dollar is written off; there is no returns platform to connect, so that gap is a limit to state), Klaviyo not connected, thin history. |
get_brief | One call for the open question: 'how are we doing', 'what should I look at this week', 'weekly review', 'where do I start'. |
get_findings | What the latest scoring run found about this store: at most three findings, each a figure, one sentence and one action, chosen by per-store thresholds in a fixed priority order: refunds at 15% of sales or more, a trade account in the top 20 by Forward CLV, 30% or more of customers leaning heavily on discounts, the top 10% holding 70% or more of Forward CLV, first products at least twice apart in value per customer, 100 or more customers due to reorder, first orders 20% or more below the eight-week pace, then the largest ranked opportunity. |
list_plays | The store's governed plays: a play is one proposed send to a named audience with a share held back, which the retailer's admin approves before anything leaves. |
get_play | One governed play in full: its audience and how it was picked, the share held back, the pre-registered measurement (primary metric and horizon), the break-even and detectable effect lines, the figures frozen at approval, the approval itself, the send receipt, and what this caller is permitted to do with it. |
get_play_results | One play's read against its own held-back group: the two arms with their assigned, delivered and converted counts, the difference in conversion with its range, every weekly look so far, the incremental profit with its range, the derivations behind each figure, and the verdict once there is one, with its reason. |
get_results | The cumulative read: everyone Tacet may contact, against the persistent held-back group, across every play. |
get_ledger | The account's governance ledger, newest first: every state a play moved through and every event the organization recorded, each with who did it, in what role, from which surface, when, the reason they gave, and the policy version in force at the time. |
get_policy | The rules this store's plays are held to: the version in force with its plain-language sentences, the clauses behind them (the share held back and its floor, quiet hours, blackout dates, the undo window, the delivery floor), who accepted it and when, any draft awaiting acceptance, and which clauses the retailer's own admin may change. |
get_capabilities | What Tacet can and cannot do in this account, read from the scopes the store's integrations actually granted rather than from what the app asks for: the Klaviyo and Shopify connections with their granted scopes, whether a reauthorization is needed before a play can run, a can list of each thing Tacet might do with the grants it requires and the ones missing, and a cannot list of true sentences about what is out of reach. |
get_holdout | The persistent held-back group every cumulative result is read against: how many customers and how many email profiles who have not bought yet, the share of the base, the share and floor the policy sets, when it was first drawn and last refreshed, and how many of its members have ever been in a play's treatment arm, which must be zero. |
get_customer_pnl | The customer P&L for a period: did the store's customers make it money, and which ones. |
get_tracker | The store's Tracker, the daily sheet on the home page of the Tacet app: one row per figure and one value per store day, with the month so far. |
get_program | The rules of the store's reminder program as they stand now. |
preview_program | What the program's next nightly run would do on the latest scores, worked out now and changing nothing: customers in each stage, how many each exclusion removed, how many would be emailed and how many held back, how many already entered this purchase cycle, how many are waiting to be sent under the daily cap, and the product each event would feature, counted by its Shopify status (active, draft, archived, not in the catalog) and by whether the store's default product would stand in. |
What Claude sees
- Scores and aggregates for your store, and customer-level scores by customer id. Email addresses are returned only when you ask for them explicitly.
- Nothing from other stores. Each request carries your own sign-in and is answered for that store only.
- No writes. The connector cannot change anything in Shopify, Klaviyo or Tacet.
Details are in the privacy statement.
Limits and support
Each store may make 60 requests a minute and 2,000 a day. Questions and problems: support@tacet.ai.
Checking service status…