Claude Code Chat downloads statistics

Claude Code for JetBrains IDE

Bring the full power of Claude Code directly into your IDE. Chat with Claude in a side panel, edit code inline with natural-language instructions, fix errors with one click in the Alt+Enter menu, generate commit messages from your staged diff, and let Claude read, edit, and create files in your project — all without leaving IntelliJ or your favorite IDE.

Inline AI Actions

• Edit with Claude (Cmd+Alt+K / Ctrl+Alt+K, or the Claude icon in the floating code toolbar) — select code, describe a change in plain English, review the diff, and accept or reject. • Fix with Claude — Alt+Enter on any IDE-detected error or warning to get an AI-powered quick-fix that lives next to IntelliJ's own intentions. • Generate Commit Message — one click in the VCS commit dialog turns your staged diff into a Conventional Commits-style message. • Apply link on every chat code block — replace the selection in your active editor with the suggested code after a diff review.

Chat Panel

• Multiple parallel Claude Code sessions in a side panel • Project-aware: Claude works in your open project's directory • Session memory: Claude remembers context across messages • Resume past chats — a clock-icon dropdown in the tool window header lists recent sessions per project. Pick one and the previous conversation auto-loads inline (read straight from Claude's local transcript, no API cost) so you can continue from where you left off after closing the IDE. Inline pencil / trash icons rename or remove entries. • Markdown rendering with syntax-highlighted code blocks • Tool-use indicators with colored status signals and diff summaries • File chips — drag/drop, copy/paste, or type a filename to attach files as context; resolves bare filenames against your project • Image paste — paste a screenshot or any clipboard image; saved to a temp PNG and sent to Claude as a multimodal input on submit • Large-text paste chips — long pastes collapse into a compact chip; double-click to expand back into the input • Queue messages while Claude works — type a follow-up (or /btw <message>) while a response is streaming and it queues instead of being lost, then sends automatically when the turn finishes; remove queued items before they go • Smart auto-scroll — follows new output unless you've scrolled up to read; "Jump to latest" button keeps you in control • Inline diffs: expandable old vs. new for every edit Claude makes, with consolidated stats per file • Clickable file links and a changed-files summary grouped by folder • Copy and re-edit buttons on messages and code blocks

Right-Click Context Menu

• Explain — understand code, methods, files, or entire folders (recursive summaries) • Refactor — get suggestions for readability, performance, and best practices • Generate Tests — create unit tests for selected code, whole files, or every source file in a folder (14+ languages) • Fix — identify and fix bugs in selected code • Open Claude Session Here — right-click any folder to start a new session rooted there • Debug with Claude — right-click in the Run/Debug console to send exceptions, log flows, or output to Claude

Works from the editor, project tree, and Run/Debug console.

Productivity

• Auto-updating model catalog — the model dropdown refreshes every 24h from a published list (with bundled offline fallback), so new Claude releases become available without a plugin update. Deprecated models get a (deprecated) badge and a one-click swap to the recommended replacement. • Keyboard shortcuts: Cmd+Shift+I to toggle the chat panel, Cmd+Alt+K for inline edit (Ctrl variants on Windows/Linux) • Stop requests with Escape • Permission mode: Plan (read-only Read/Grep/Glob), Content Only (file writes/edits OK, shell blocked), or Unrestricted (allow all) — maps to the CLI's --permission-mode • Project tree auto-refreshes when Claude creates, modifies, or deletes files • Cost tracking per response

Settings — Model & inference

• Extended thinking budget — Off / Low (1k) / Medium (8k) / High (32k). Sets MAX_THINKING_TOKENS for Opus/Sonnet 4.x; other models ignore it. Trade speed and cost for deeper reasoning on harder tasks. • Append system prompt — multi-line text appended to Claude's system prompt every spawn via --append-system-prompt. Use for project conventions, preferred tone, persona instructions. • Max agentic turns — hard cap on the tool-call loop via --max-turns. 0 = unlimited (CLI default).

MCP Servers

• MCP server management UI — a server icon in the tool-window header (and a button in Settings) opens a manager that aggregates every MCP server Claude can see for the project: project .mcp.json, plus the user and project-local scopes inside ~/.claude.json. Add, edit, and remove servers across all three scopes. • All transports — stdio (command + args + env) and remote SSE / HTTP (URL + headers, optional OAuth client id / callback port). • Live status — connection/auth badges (✓ connected, ✗ failed, ⏸ pending approval, 🔒 needs authentication) from a health-checked claude mcp list. • Authentication, your way — choose API Key / Token, OAuth, or No Auth per server. OAuth sign-in runs in the background and simply opens your browser to finish (no terminal); Authenticate All walks every server that needs it, one at a time, with a live progress panel (skip/remove, countdown). Your chat sessions then use the servers automatically. • All writes go through the claude mcp add-json / remove CLI, so your large, live ~/.claude.json is never hand-edited.

Settings Files Editor

• Edit Claude Code's settings.json at project, project-local, or global scope — a JSON-file icon in the header (and a Settings button) opens it. • Structured or raw — a key/value Form tab adds/edits/removes top-level settings (with common keys suggested and typed JSON values), or switch to the JSON tab for raw edits. Both are validated as a JSON object before saving, and unrelated keys (including the plugin's permission grants) are preserved.…

Description from the JetBrains Marketplace listing.