Where each agent runs
Junie is JetBrains' coding agent. Junie's site leads with Junie CLI, which runs on Linux, macOS and Windows, and the same agent works in JetBrains IDEs, in GitHub through a GitHub Action, in GitLab CI/CD, and across projects in Air in IDEs. Inside the IDE, JetBrains recommends selecting Junie in the AI Chat tool window; a separate Junie plugin is also available for IntelliJ IDEA Ultimate, PyCharm Pro, WebStorm, GoLand and other JetBrains IDEs, and Junie runs in Android Studio.
Claude Code is Anthropic's agentic coding tool. Its documentation lists the terminal CLI, a VS Code extension (also installable in Cursor), a JetBrains plugin that needs the CLI installed separately, a desktop app for macOS and Windows with a Linux beta on Ubuntu and Debian, and a web version at claude.ai/code. Anthropic also documents GitHub Actions and GitLab CI/CD integrations.
Models and billing
The two tools take opposite approaches to models. Claude Code runs Anthropic's models: on the Anthropic API, its opus, sonnet and haiku aliases resolve to Opus 5.5, Sonnet 5.5 and Haiku 5.5, and the same aliases work through Amazon Bedrock, Google Cloud and Microsoft Foundry.
Junie's site calls it a coding agent that works with any model you choose. Its product page lists Claude Opus 5.5, Claude Sonnet 5.5, Claude Fable 5.1, Gemini 3.1 Pro, Gemini 3.8 Flash, GPT-6 Astra, GPT-6.1 Sol and Grok 4.7 among supported models, and it suggests planning on a powerful model and implementing on a faster one. Junie CLI can authenticate in three ways: a JetBrains account with a JetBrains AI subscription, a Junie API key with usage-based billing, or your own API keys from OpenAI, Anthropic, Google, xAI, OpenRouter or GitHub Copilot. It also connects to custom or self-hosted endpoints such as LiteLLM, Ollama and LM Studio. When a model is available through both BYOK and a JetBrains subscription, Junie bills the request to your BYOK provider.
Pricing
Claude Code requires a paid Claude plan or an API key. Pro costs $20 per month ($17 per month billed annually), Max costs $100 for 5x Pro usage or $200 for 20x, Team seats cost $25 (Standard) or $125 (Premium) per month, or $20 and $100 billed annually, and Enterprise is $20 per seat per month billed annually with a 20-seat minimum plus usage at API rates.
Junie is paid through JetBrains AI credits, where one credit equals $1. For individuals, AI Pro costs $10 per month with 10 AI credits per 30 days, and AI Ultimate costs $30 per month with 35 credits; JetBrains recommends Ultimate for regular Junie use. On annual billing, Junie's pricing section shows $8.33 and $25.00 per user per month. AI Free includes 3 credits per 30 days. Extra credits can be bought at any time and stay valid for 12 months. Organization prices are higher: AI Pro is $20, AI Ultimate $60 and AI Enterprise $60 per user per month. Junie's site also advertises Junie Lite, a free model available in every JetBrains account, and BYOK at provider rates with no markup.
Control and safety
Both agents ask before risky actions and let you loosen that. Junie CLI asks for approval before most terminal commands, edits outside the project and MCP tool calls. You can add commands to an Action Allowlist, stored in ~/.junie/allowlist.json, and cycle Brave mode between Off, Auto (a safety check auto-approves commands it considers safe) and On. On macOS and Linux, a sandbox mode (currently in JetBrains' Early Access Program) restricts agent terminal commands and file edits. JetBrains notes that MCP servers and hooks are not isolated by this sandbox.
Junie's plan mode analyzes the codebase with read-only operations and writes a design document before any code changes. Live prompting lets you add instructions while a task is running, and the /review command reviews local changes with the same backend JetBrains uses for automated pull request reviews. Debug mode can drive breakpoints and inspect runtime state in a connected JetBrains IDE, which is the clearest IDE-specific advantage Junie has.
Claude Code supports MCP servers, CLAUDE.md project instructions (and can read an existing AGENTS.md), skills, hooks, subagents and background agents. It also offers cloud and desktop scheduled tasks and an Agent SDK for building custom agents on Claude Code's tools. Junie keeps its project guidelines in .junie/AGENTS.md and supports MCP, custom slash commands and Agent Skills packages.
Who should pick which
- Pick Junie if your team works in IntelliJ-based IDEs, wants an agent that can use the IDE's debugger, or wants to switch between Claude, GPT, Gemini and Grok models under one JetBrains AI subscription or your own keys.
- Pick Claude Code if you want Anthropic's models with a CLI-first workflow that also runs in VS Code, a desktop app and the browser, or if you need to route traffic through Bedrock, Google Cloud or Foundry.
- If you already pay for JetBrains All Products Pack or AI Pro, start with Junie. If you already pay for Claude Pro or Max, start with Claude Code.
The Bottom Line
Claude Code and Junie overlap more than they used to: both are terminal agents with plan modes, approvals, MCP and CI integrations. The deciding factors are models and home base. Claude Code ties you to Claude models and gives you the widest set of Anthropic surfaces. Junie lets you pick models from several vendors, adds JetBrains IDE features such as debugger control, and bills through JetBrains AI credits or your own keys.
