Set up RevenueCat MCP Server
Connect the hosted RevenueCat MCP server to an AI assistant, set up an API key, and protect write actions in the production configuration.
- Skill Road
- Set up RevenueCat MCP Server
Published on 18.09.2026
The RevenueCat MCP Server connects your AI assistant directly to your RevenueCat project. The server is hosted by RevenueCat, so there is nothing to install locally — you only connect the assistant to the remote endpoint and sign in once.
Create an API key
Open your RevenueCat dashboard, go to the project settings, and switch to the "API keys" tab. Create a new API v2 secret key. Grant only the permissions you actually need: read access is enough for queries, write permissions are required if you want the assistant to create or modify products, offerings, or entitlements. Copy the key immediately — it is only shown in full once.
Connect the endpoint
Add the remote server https://mcp.revenuecat.ai/mcp in your MCP client. Clients with OAuth support (Claude, Claude Code, Codex, Cursor, VS Code, GitHub Copilot) can sign in via OAuth without an API key. For Claude Code the command is: claude mcp add --transport http revenuecat https://mcp.revenuecat.ai/mcp --header "Authorization: Bearer YOUR_KEY". Clients without native remote support use the bridge npx mcp-remote https://mcp.revenuecat.ai/mcp --header "Authorization: Bearer YOUR_KEY".
RevenueCat AI Toolkit (recommended path for coding assistants)
For Claude Code, Codex, and Cursor, RevenueCat recommends the AI Toolkit instead of manual setup — a plugin that automatically configures the MCP server and also includes pre-built workflows (skills) for common RevenueCat tasks.
Run a first test
Start with a read-only query, such as "Show me all products in my project" or "List the offerings for my iOS app". Check whether the response matches the dashboard. Then move to more complex queries or — with write permissions in place — try first write actions.
Protect write actions
Products, offerings, and entitlements take effect directly on your app's production configuration. Always have the assistant describe changes before it executes them. For experimental workflows it is worth working in a development or test environment first and using a separate API key with limited permissions.
Frequently asked questions
Do I need to self-host the server?
No. RevenueCat operates it fully hosted at https://mcp.revenuecat.ai/mcp. There is no public repository for operating the server itself; only the AI Toolkit plugin is open source.
Which RevenueCat plan do I need?
According to the provider, the MCP server is available to all RevenueCat accounts, including the free starter plan. Feature scope and API limits follow the subscribed plan; current amounts are only on the official pricing page.
Can I use OAuth instead of an API key?
Yes. OAuth is supported for Claude, Claude Code, Codex, Cursor, VS Code, and GitHub Copilot and is the preferred method because no manual key needs to be managed. For older or non-OAuth clients the Bearer token approach remains available.
Can the server retrieve customer data of my app users?
Yes, depending on the permissions of the API key in use, customer data can be viewed. Dedicated keys with a minimal permission profile — for example read-only access for the given project — are recommended to limit the risk of unintended data access.
What is the difference between the MCP Server and the AI Toolkit?
The MCP server provides the tools that let an assistant read and write RevenueCat data. The AI Toolkit is a plugin for Claude Code, Codex, and Cursor that automatically sets up the MCP server and also includes pre-built workflows (skills) for common RevenueCat tasks. For simple clients or other assistants, the MCP server on its own is sufficient.