AppActor docs are structured so AI tools can quickly find the right SDK, concept, or recovery page before answering implementation questions.

​
Connect AppActor to your AI client

AppActor runs a remote MCP server, so an assistant can read your workspace and change your catalog directly instead of being told about it. Both clients below install from the same place, appactor/appactor-aikit, which carries the server connection and the skills together.
In Claude Code, run /mcp, pick appactor, and choose Authenticate if you are not prompted on the first tool call. Codex signs in as part of mcp login. Any other MCP client can take the server URL on its own:
Every client discovers AppActor’s OAuth automatically and opens a browser to sign in, where you choose which organizations and projects the connection may reach. The AppActor session cookie never leaves the browser, and your existing AppActor permissions still apply to every request. A connection stays signed in until you revoke it, from Settings → AI connections in the dashboard. Read tools: get_workspace, get_app_setup, query_analytics, get_catalog, get_subscriber, get_config, get_audit_log. Write tools: manage_products, manage_entitlements, manage_offerings, manage_packages, manage_remote_config, manage_experiments, create_project, create_app. Deletion, entitlement detach, credential handling, and key rotation are intentionally unavailable to AI clients and stay in the dashboard.

​
Skills

Installing the plugin adds skills for the Flutter, iOS, Android, and React Native SDKs, plus paywall and catalog modelling, remote config and experiments, and troubleshooting. Claude Code and Codex both read them. The SDK skills are written from the SDK sources, so signatures, enum cases, error codes, and ordering constraints match the code rather than a summary of it. Registering only the server URL gives you the tools without the skills. For SDK implementation questions, load:
  1. How AppActor Works
  2. Project and Store Setup
  3. Catalog Setup
  4. Server Notifications
  5. Outbound Webhooks, if the user’s backend reacts asynchronously
  6. Entitlements
  7. Products and Packages
  8. Offerings
  9. Displaying Products
  10. Making Purchases
  11. Trusted Entitlements
  12. the relevant SDK installation and quickstart page
  13. the matching lifecycle, purchase, and recovery pages
For product-surface questions, load the product model before platform code:

​
Machine-readable index

Mintlify serves /llms.txt and /llms-full.txt for this site. Point a crawler or an agent at those rather than scraping the rendered pages.

​
Human review still matters

AI-generated integration guidance should be checked against the SDK version you ship and the store behavior you are testing.