Use subscription data inside IntelliJ with an AI assistant
Developers often leave their editor to check revenue numbers, confirm paywall identifiers, or verify integration settings. An in-IDE assistant connected to subscription tooling can shorten that loop: ask a question in a chat panel, get a formatted answer, and return to coding. This guide explains how that workflow is set up, which questions it handles well, and how to pick a model for the task at hand.
The foundation: MCP tools for subscription data
The Model Context Protocol is an open standard that lets AI assistants work with external tools and data. The subscription platform's MCP Server applies that pattern to subscription management, exposing tools for metrics, offerings, packages, products, customer subscription information, and documentation links. The IntelliJ plugin's assistant uses the same underlying tool architecture, so answers inside the IDE draw on the same capabilities rather than a separate, inconsistent data path.
Setup checklist
Setup follows a short checklist:
- Open IntelliJ Settings and go to the subscription platform's settings, then AI Settings.
- Turn on the AI Assistant toggle.
- Pick a provider and model from the supported list.
- Enter the provider API key, which is kept in IntelliJ's credential storage on the local computer.
- Confirm the platform API credentials are set in the main settings panel.
After that, the assistant appears as a chat-style panel in the platform tool window, where you type a question and read the response in place.
What to ask: four core capabilities
Four kinds of questions cover most day-to-day work.
Fetching subscription metrics: Ask for current metrics or MRR to see monthly recurring revenue, active trials, active subscriptions, and total revenue in a formatted reply. Follow-ups about trial conversion, revenue growth, or ideas to improve revenue let the assistant reason over the numbers you just retrieved. A practical pattern is to ship a paywall change, then immediately check whether metrics move in the expected direction without opening a dashboard.
Exploring offerings and packages: Ask what offerings are configured to get lookup keys, display names, current-offering status, and the packages inside each offering. That is useful when writing purchase code and you need exact identifiers, or when you want SDK code guidance for a platform such as fetching offerings on Android with Kotlin.
Checking project configuration: Ask for project status to see whether the API key is configured, the project ID, SDK API key setup, and notification and webhook status. Treat this as a pre-flight check before implementation so a missing setting is caught before it becomes a runtime bug.
Finding documentation and resources: Ask where to find Android integration docs or how to set up webhooks to receive curated links covering documentation, SDK guides for Android, iOS, Flutter, and Kotlin Multiplatform, community forums, GitHub issues, the dashboard, and the API reference.
Choosing a model by task
Match the model to the question rather than defaulting to one choice. For complex analysis and detailed explanations, GPT-4o and Claude 3.5 Sonnet are the stronger options, for example when asking for trend analysis and improvement suggestions. For quick lookups such as MRR or a list of offerings, GPT-4o Mini, Claude 3 Haiku, and Gemini 2.5 Flash respond faster and cost less per query. Gemini 2.5 Pro sits in the middle, with strong reasoning and good response times. Switching models in settings lets you test which fits your workflow.
From answers to decisions
The same answers support three higher-level jobs. First, optimization: ask what the metrics say about subscription health, then review patterns such as trial conversion or churn and consider responses like reviewing onboarding or adding win-back campaigns. Second, experiment validation: list all offerings and their packages to confirm test and control offerings are configured before an experiment runs. Third, purchase debugging: check project configuration for an instant health check, and inspect the default offering's packages to confirm users should see the right products. A help icon in the assistant links to MCP usage examples with more query patterns, and the plugin's metrics dashboard, offerings explorer, SDK release notifications, and webhook listener complement the chat panel.
Bottom line
The value is reduced context switching: metrics, offerings, configuration, and documentation are reachable from the editor, so development cycles stay in flow. Configure the AI provider, connect the platform credentials, and start with simple lookups before moving to analysis questions.
Frequently asked questions
How do I set up the AI Assistant in the IntelliJ plugin?
Open IntelliJ Settings, go to the subscription platform's settings and then AI Settings, enable the AI Assistant toggle, select a provider and model, enter that provider's API key, and confirm the platform API credentials are configured in the main settings panel. The provider key is stored in IntelliJ's local credential storage.
Which AI model should I choose?
Use GPT-4o or Claude 3.5 Sonnet for complex analysis and detailed explanations. Use GPT-4o Mini, Claude 3 Haiku, or Gemini 2.5 Flash for quick lookups and faster, lower-cost responses. Gemini 2.5 Pro is presented as a balanced option with strong reasoning and good response times. You can switch models in settings.
What can the assistant answer from inside the IDE?
You can ask for current metrics such as MRR, active trials, active subscriptions and total revenue; list offerings with lookup keys, display names, current-offering status and their packages; check API key, project ID, SDK key, notifications and webhooks status; and get links to documentation, SDK guides, forums, GitHub issues, the dashboard and API reference.