MCP tools reference

The RevenueHog MCP server registers eight tools. Each one is scoped to the workspace the credential belongs to, and each returns a single text block of JSON. On failure the block is {"error": "<message>"} and the result has isError: true.

rh_queryCopy Anchor Link

The main analytics tool. It crosses one or more measures with an optional dimension, over a window, narrowed by filters.

ArgumentsCopy Anchor Link

WindowsCopy Anchor Link

All windows are in UTC and end today (UTC midnight).

ReturnsCopy Anchor Link

{
  "window": { "from": "2026-09-28", "to": "2026-10-28", "days": 31 },
  "compareTo": { "from": "2026-08-28", "to": "2026-09-27" },
  "currency": "USD",
  "measures": [{ "id": "proceeds", "label": "Proceeds", "unit": "usd" }],
  "rows": [{ "app": "app_123", "proceeds": 1840.22, "proceeds_delta": 0.12 }],
  "notes": ["Prior-period comparison is unavailable for an all-time window."]
}

• Without a dimension, rows has one totals row.
• With date, each row has date plus one column per measure, sorted by date.
• With a categorical dimension, each row has the dimension key (for app, the RevenueHog app ID) plus one column per measure.
• <measure>_delta is a fraction (0.12 means up 12%). It is null when either side is unknown or there is no positive baseline.
• notes appears only when there is something to say.

Measure catalogCopy Anchor Link

Formulas for every measure are on Metric definitions, and rh_meta with topic glossary returns them at runtime.

ErrorsCopy Anchor Link

rh_metaCopy Anchor Link

Discovery and reference for rh_query.

lastData is the newest day of data, which differs from when a sync last ran. Apple sales reports settle 1 to 3 days late, the subscription report lags 1 to 2 days, and Apple Ads about 1 day.

rh_reviewsCopy Anchor Link

Browse mirrored App Store customer reviews, newest first.

Returns { reviews, nextCursor, trend }:

• Each review: id, appId, rating, title, body, reviewerNickname, territory, createdDate, hasResponse, responseBody, responseState, responseIsAi, responseUpdatedAt.
• nextCursor is null on the last page.
• trend is { avg, count, byApp: { <appId>: { avg, count } } }. avg is count-weighted and rounded to one decimal, or null with no reviews. It respects the app filter but not the other filters.

Error: No app found matching "x" in this workspace.

rh_feedbackCopy Anchor Link

Browse in-app feedback from the SDK, the feedback API, or manual entry, newest first.

Returns { feedback, nextCursor, summary }:

• Each item: id, kind, message, status, source, platform, rating, test, attested, unread, createdAt, appId, appName, appIconUrl, customer.
• customer is { id, handle, attested, plan, ltvUsd } when the write identified one, otherwise null.
• summary is { unread, total, byKind }, excluding test feedback, scoped only by app.

Error: No app found matching "x" in this workspace.

rh_experiment_resultsCopy Anchor Link

Paywall A/B experiments. Two modes, chosen by experimentId.

List mode returns { mode: "list", filters, count, experiments }. Each experiment has id, appId, app, entitlement, name, status, effectiveStatus, durationDays, startedAt, endsAt, targetExposures, concludedVariant, createdAt, totalExposures and variants (key, isControl, weightPct, skus, assigned, exposed). Counts use production assignments only.

Results mode returns { mode: "results", experiment, results, analysis }:

• results is the full computed result set used by the dashboard, including health and the sample ratio mismatch check.
• analysis.variants gives, per variant: exposures, conversions (paid installs), conversionRatePct, paidRate (with Wilson 95% bounds), realizedPerExposureUsd, normalizedPerExposureUsd, relativeLift (fraction versus control), absoluteLiftPct (percentage points) and significantVsControl (true when the Wilson intervals do not overlap).
• analysis also has controlKey, concludedWinner, leadingSignificantVariant and srmFlagged.
• An experiment that has not started returns results: null and the note "This experiment has not started, so there are no results yet."

rh_payoutsCopy Anchor Link

What Apple actually pays, from monthly financial reports.

Returns { periods, nextExpectedPayment }, newest first. Each period has period, startsAt, endsAt, expectedPaymentAt, status, earnedUsd, units, byCurrency (currency, earned), byApp (appId, name, earnedUsd; app-less lines grouped as "Other"), estimatedUsd (Sales and Trends proceeds over the same dates) and variancePct. nextExpectedPayment is { period, status, expectedPaymentAt, amountUsd } for the soonest unpaid period, or null.

rh_create_api_keyCopy Anchor Link

Create a server-to-server feedback API key and return its secret once. Apps using the RevenueHog SDK do not need one.

Returns { record, secret, warning }. secret is the full key and is never retrievable again. record is the saved summary (see below). Revoke keys in the dashboard under Settings → Connections → API KEYS; the MCP server cannot revoke.

rh_list_api_keysCopy Anchor Link

No arguments. Returns { keys }, newest first, never including secrets. Each key has id, name, prefix, mode, appId (null for workspace-wide), scopes, lastUsedAt, revokedAt and createdAt. Revoked keys are included; a non-null revokedAt means the key no longer works.

RelatedCopy Anchor Link