RevenueCat MCP: RevenueCat's server, and app revenue next to ad spend

Yes. RevenueCat hosts an MCP server at https://mcp.revenuecat.ai/mcp, part of its AI Toolkit alongside plugins, skills and a CLI. RevenueCat MCP gives a coding agent a live connection to your RevenueCat projects: the docs list project details, app management, products and in-app purchases, store product state such as prices and availability, offerings and packages, paywalls, and an analytics side for charts and experiment results.

Sign-in is OAuth with your RevenueCat account on Claude, Codex, Copilot, Cursor, VS Code and several other clients, so the agent gets the access your account has across your projects. Every MCP client can instead send an API v2 secret key as an Authorization Bearer header; RevenueCat advises a dedicated key for the server, write-enabled when the agent should change things and read-only when it should only look.

The documented setups are Claude Code with claude mcp add --transport http revenuecat https://mcp.revenuecat.ai/mcp, Codex through the RevenueCat plugin marketplace, Cursor with /add-plugin revenuecat or an mcp.json entry, VS Code Copilot with the same URL, and Claude on the web or desktop as a custom connector. The toolkit's plugins also name Gemini CLI. RevenueCat's pages state no plan requirement and no price for the server.

This is a configuration server first. It creates, reads, updates and deletes apps, manages products, offerings and packages, changes store product prices, availability and review metadata, and creates and updates paywalls. A developer wiring a new subscription tier or a paywall test has every lever in the conversation, which is why the key you give it decides what the agent may touch.

The analytics side reads charts and experiment results. Customers, subscriptions, entitlements, purchases and invoices are objects of RevenueCat's REST API v2 at https://api.revenuecat.com/v2, read with a v2 secret key whose permissions you set at creation, under rate limits by domain: 480 requests a minute for customer information, 60 for project configuration, 25 for charts and metrics. A secret key can grant entitlements and delete subscribers, so a key with write permissions belongs on a server, not in a chat client.

What RevenueCat's server does not hold is the ad account. It can say that MRR moved this month and which experiment variant converted; it cannot say what the subscribers of that month cost on Meta or Google Ads, by campaign, because the spend is not in RevenueCat.

• Writes: apps, products, offerings, packages, paywalls, store product state
• Reads: project details, charts, experiment results
• REST API v2: customers, subscriptions, entitlements, purchases, invoices, with per-domain rate limits

Adrails connects RevenueCat as a data source. In RevenueCat you create a secret API key, version 2, for your project; in Adrails you open Integrations, choose RevenueCat and paste the sk_ key. Adrails finds the project from the key and uses four read permissions: project_configuration:projects:read for the project, charts_metrics:overview:read for revenue, MRR and the overview metrics, customer_information:customers:read for customers, and customer_information:subscriptions:read for subscriptions and their billing cadence.

Subscriptions are read every 30 minutes, revenue metrics and customers every 6 hours. The key only reads, and Adrails never changes anything in RevenueCat: no product, no offering, no paywall. Those stay with RevenueCat's own server or its dashboard.

An AI client connected to https://adrails.ai/api/mcp reads them through saas_performance: for a period, Stripe billing, RevenueCat revenue metrics and PostHog product events, with active subscriptions, trials, normalized MRR, ARR, collected payments and refunds, and customer and subscription counts by billing cadence when customer-level data is available. Each source and currency stays separate; nothing is converted or merged into the platform's own revenue.

Cost per install paces an app campaign; cost per paying user decides it. Adrails holds both sides. performance_period returns one ad account's spend, results, CPA and ROAS for exact dates at the campaign, ad set or ad level, with a comparison range; saas_performance returns the trials, active subscriptions and MRR RevenueCat recorded over the same dates. planning_calculate divides one by the other as first-order acquisition unit economics, in one currency, with the sources and periods stated, and returns the arithmetic rather than a forecast.

business_review checks a calculation the assistant or you wrote, such as a payback period, and refuses any number that does not quote its source word for word from the conversation's own tool results. Ask what a paying subscriber cost on Meta last month: the answer names the spend, the new paying customers, their periods, and the point where the data stops.

In Adrails itself, a campaign's Tracking tab lists the customers each ad brought and their monthly revenue when the journey carried the campaign and ad ids, as a web funnel does; for a store-only funnel, Adrails sets RevenueCat totals next to spend for the same period rather than claiming a per-ad match it cannot see. The agent reads revenue, MRR, customers and subscriptions by billing cadence, a weekly review automation with an agent step reads the campaigns and RevenueCat together and emails the ads worth a look, and a CPA alert checks cost per result every hour.

• Spend by campaign for exact dates: performance_period
• Trials, active subscriptions, MRR by billing cadence: saas_performance
• Cost per trial and per paying user, sources stated: planning_calculate

Connect RevenueCat's server when the work is in RevenueCat: a new product, an offering, a paywall, a look at a chart or an experiment while coding. It is the vendor's complete configuration surface, it writes, and it signs in with your account's access across every project, so match the key or the OAuth session to what the agent should be allowed to do.

Connect Adrails when the question has ad spend in it: cost per paying user by campaign, whether a renewal cohort justifies the budget raise that preceded it, which campaign to pause, what the weekly review should say. Adrails never writes to RevenueCat, and it brings Meta Ads and, once your workspace can connect it, Google Ads, together with Stripe, PostHog and the store sources, into the same conversation. Every advertising change is a proposal you apply in Adrails, and a budget move is capped at 30% per change.

Both at once is the natural setup in Claude or Claude Code: the servers do not overlap. Ask RevenueCat's server why an offering converts, ask Adrails what the campaign that fed it spent, and let the assistant hold both answers. Each server signs in on its own account and keeps its own permissions.

For RevenueCat's server, Claude Code takes claude mcp add --transport http revenuecat https://mcp.revenuecat.ai/mcp; Claude on the web or desktop takes the same URL as a custom connector; Cursor takes /add-plugin revenuecat; Codex adds the RevenueCat plugin marketplace, installs the plugin and runs codex mcp login RevenueCat. Sign in with OAuth, or paste a dedicated API v2 key as the Bearer header.

For Adrails, add https://adrails.ai/api/mcp as a custom connector in Claude under Customize, then Connectors; run claude mcp add --transport http adrails https://adrails.ai/api/mcp in Claude Code; in ChatGPT, turn on Developer mode and add a custom MCP server with OAuth as its authentication. Sign in to Adrails, choose the workspace and the permissions, then allow access. Read access is always included, write permissions only if you select them, and one connection works in one workspace.

Then connect RevenueCat inside Adrails, under Integrations, with the version 2 secret key and the four read permissions, or ask the assistant for a setup link with integrations_connect, valid for 10 minutes and opened by an owner or admin. Ask which integrations are connected: integrations_status lists RevenueCat and how fresh each of its data streams is, and the first saas_performance call returns MRR, trials and active subscriptions next to the spend of the same period.

• RevenueCat's server: the URL plus OAuth, or a dedicated API v2 key
• Adrails: https://adrails.ai/api/mcp as a custom connector, OAuth sign-in
• RevenueCat in Adrails: Integrations, an sk_ key with four read permissions