Unlimited Claude in JetBrains IDEs

JetBrains IDEs — IntelliJ IDEA, PyCharm, WebStorm, GoLand, and the rest — support Claude through the Continue plugin available from the JetBrains Marketplace. Continue uses a shared configuration file that works identically across all JetBrains products, so you configure once and get Claude in every IDE you use. Connecting through an unlimited gateway means your AI assistant is always available across all your IDEs without per-token billing accumulating from each one.

Installing Continue in Any JetBrains IDE

Open your JetBrains IDE and navigate to Settings (Ctrl+Alt+S) then Plugins. Search for Continue in the Marketplace tab and click Install. The plugin works in all JetBrains IDEs from version 2023.1 onward — IntelliJ IDEA, PyCharm, WebStorm, GoLand, PhpStorm, RubyMine, CLion, Rider, and DataGrip. One installation per IDE, but they all share the same configuration file.

After installation, restart the IDE and you will see the Continue panel in the right sidebar. The first launch creates the configuration directory at ~/.continue/ with a default config.yaml file. This is the single file you need to edit to connect Claude through your unlimited gateway. The same file is read by all JetBrains IDEs and VS Code if you use Continue there too.

The Continue plugin provides three main features: Chat (conversational AI in the sidebar), Edit (inline code modifications), and Apply (applying suggested changes to files). Each feature can be assigned to a different model, giving you fine-grained control over which Claude variant handles which task.

Configuring config.yaml for Unlimited Claude

Open ~/.continue/config.yaml in any text editor. The file uses YAML format with a models array that defines available AI providers. Add a new model entry with provider set to anthropic, apiBase set to https://unlimited.aiprimetech.io (root host, no /v1 — the SDK appends the path), and apiKey set to your AI Prime Tech Unlimited key.

Each model entry needs a title (display name in the UI), model (the model identifier), and the provider configuration. Create entries for each Claude variant you want available: one for Sonnet as your everyday model, one for Opus for complex reasoning, and optionally one for Haiku for quick interactions. The title field is what appears in the Continue model picker.

You can also use the openai provider type instead of anthropic. In that case, set apiBase to https://unlimited.aiprimetech.io/v1 (with /v1) and the key goes in the apiKey field with Bearer authentication. The anthropic provider type is recommended because it supports native features like thinking blocks and has better streaming behavior in Continue.

Role Assignments: Chat, Edit, and Apply

Continue lets you assign different models to different roles. The chat role handles sidebar conversations — code explanations, architecture discussions, debugging help. The edit role handles inline code modifications triggered by Ctrl+I — generating replacements for selected code. The apply role handles taking generated code and merging it into existing files.

In config.yaml, set roles for each model using the roles array: roles: [chat, edit, apply]. A model with all three roles is used for everything. To split responsibilities, create multiple model entries with different role assignments. For example, assign Opus to the chat role for deep reasoning conversations and Sonnet to edit and apply roles for fast inline modifications.

On unlimited, the optimal strategy is to assign Opus to chat (where you ask complex questions and need thorough analysis) and Sonnet to edit and apply (where speed matters because you are waiting for inline results). Both are covered at the same flat rate, so the choice is purely about latency and quality tradeoffs for each interaction type.

Per-Project Overrides with .continuerc.json

While config.yaml is global, you can override settings per-project using a .continuerc.json file in your project root. This is useful when different projects need different model configurations — perhaps a Python project benefits from specific system prompt additions, or a legacy codebase works better with a particular model version.

The .continuerc.json file uses the same structure as config.yaml but in JSON format. Any settings specified in the project file override the global config for that project only. Common overrides include custom system prompts (to inject project-specific coding conventions), model selections (to use a specific model version for a particular codebase), and context providers (to add project-specific documentation sources).

On unlimited, per-project overrides let you tune the AI experience for each codebase without cost implications. Add verbose system prompts with full architecture documentation for complex projects. Use Opus by default for challenging legacy codebases. Use Haiku for projects where you mainly need quick lookups. The flat rate covers all configurations equally.

Context Providers and Codebase Indexing

Continue supports context providers that inject additional information into requests. The built-in codebase provider indexes your project files and includes relevant code snippets when you ask questions. The docs provider can index documentation URLs. The terminal provider includes recent terminal output. Each active provider adds tokens to every request.

Configure context providers in config.yaml under the contextProviders section. The codebase provider uses an embeddings model to index your project — this runs locally and does not consume API tokens. However, the retrieved context snippets are included in the prompt sent to Claude, adding to the per-request token count. With a large codebase index, this can add 5,000-15,000 tokens per request.

On unlimited, enable all context providers that improve response quality. The more relevant context Claude has, the better its answers. Add documentation providers for your framework, API references for services you use, and custom providers for internal knowledge bases. The extra tokens per request are covered by the flat rate and the quality improvement justifies the overhead.

Cross-IDE Consistency and Shared Config

One of Continue's strongest features is configuration portability. The ~/.continue/ directory is shared across all IDEs that use Continue — every JetBrains product plus VS Code. Edit config.yaml once and the change applies everywhere. Your unlimited Claude connection works in IntelliJ when writing Java, in PyCharm when writing Python, in WebStorm when writing TypeScript, and in VS Code for everything else.

This shared config means your model preferences, role assignments, system prompts, and context providers are consistent regardless of which IDE you open. Developers who work across multiple languages and multiple JetBrains IDEs get a unified AI experience without repeating configuration in each tool.

The unlimited plan amplifies this benefit because cross-IDE usage multiplies token consumption. A developer who uses IntelliJ, PyCharm, and WebStorm throughout the day — asking questions, generating code, reviewing changes in each — accumulates API usage across all three. On per-token billing, this cross-IDE pattern gets expensive. On unlimited, all IDEs are covered under one flat rate.

Troubleshooting Continue in JetBrains

If Continue shows 'model not available' after configuration, check the config.yaml syntax carefully. YAML is whitespace-sensitive — incorrect indentation silently breaks model entries. Use a YAML validator or open the Continue output panel (View -> Tool Windows -> Continue) to see parsing errors. The most common issue is a missing space after colons in key-value pairs.

If requests timeout in JetBrains but work in VS Code with the same config, the issue may be the JetBrains proxy settings overriding direct connections. Check Settings -> Appearance & Behavior -> System Settings -> HTTP Proxy. If set to 'Auto-detect' or a corporate proxy, try 'No proxy' to confirm direct connectivity, then configure the proxy correctly if needed.

Authentication errors (401/403) with correct credentials sometimes indicate that the config.yaml was not reloaded after editing. Close and reopen the Continue panel, or restart the IDE. Continue caches the configuration on panel initialization and does not watch the file for changes in all IDE versions.

# ~/.continue/config.yaml
models:
  - title: "Claude Sonnet (Unlimited)"
    provider: anthropic
    model: claude-sonnet-4-5
    apiBase: "https://aiprimetech.io"
    apiKey: "<your AI Prime Tech Unlimited key>"
    roles:
      - edit
      - apply

  - title: "Claude Opus (Unlimited)"
    provider: anthropic
    model: claude-opus-4-6
    apiBase: "https://aiprimetech.io"
    apiKey: "<your AI Prime Tech Unlimited key>"
    roles:
      - chat

# This config works in ALL JetBrains IDEs + VS Code simultaneously

Frequently asked questions

Does one config.yaml work across all JetBrains IDEs?
Yes. The ~/.continue/config.yaml file is shared by all JetBrains IDEs (IntelliJ, PyCharm, WebStorm, GoLand, etc.) and VS Code. Configure once, use everywhere. Your unlimited Claude connection is available in every IDE without repeated setup.

Should I use the anthropic or openai provider type in Continue?
Use the anthropic provider type with apiBase set to the root host (no /v1). It provides native support for streaming, thinking blocks, and tool calling. The openai provider type works but requires /v1 in the apiBase and loses some Anthropic-specific features.

Can I assign different models to chat vs edit in Continue?
Yes. Create multiple model entries with different roles arrays. Assign Opus to the chat role for deep reasoning and Sonnet to edit/apply roles for fast inline modifications. On unlimited, this split optimizes for both quality and speed without cost tradeoffs.

Why does Continue add so many tokens from context providers?
Context providers (codebase index, docs, terminal) inject relevant information into every request. This can add 5,000-15,000 tokens per call but significantly improves response quality. On unlimited, enable all useful providers without worrying about the per-request token overhead.

How do I override settings for a specific project?
Create a .continuerc.json file in your project root. It uses the same structure as config.yaml (in JSON format) and overrides global settings for that project only. Useful for project-specific system prompts, model preferences, or context providers.

AI Prime Tech is an independent API gateway. It is not affiliated with, endorsed by, or a reseller of Anthropic. Claude and related model names are trademarks of their respective owners.