A desktop app to explore, manage, and work with your local Claude Code data — macOS, with experimental Linux and Windows support.

If you use Claude Code heavily, you know how opaque ~/.claude/ is. ClaudeLens makes it navigable: read your past sessions, manage memory and CLAUDE.md, edit skills and agents — and pick a conversation back up, either in an embedded terminal or in an in-app chat. Everything stays in sync with ~/.claude/ in real time, so changes you make in the terminal show up instantly, and vice versa.

Global Home — Light	Global Home — Dark
Project Overview — Light	Project Overview — Dark
Skill Editor — Light	Skill Editor — Dark
Agent View — Light	Agent View — Dark

• Projects & sessions — every project from ~/.claude/projects/ in one sidebar, with token usage and cost per project and per session. Open any session to replay the full conversation: messages, thinking blocks, and tool calls rendered for what they are (code diffs, terminal output, web results, sub-agent transcripts).
• Analytics — token and cost trends across your history.
• Memory — read and edit memory topics with full CRUD, kept in the exact MEMORY.md + topic-file format Claude Code expects, with a warning when the index grows large enough to risk truncation.
• CLAUDE.md & rules — the complete instruction hierarchy active for a project (global → project → local → subdir) plus conditional .claude/rules/.
• Skills, agents, plugins & MCP — browse global and per-project skills, agents, installed plugins, and MCP servers; create or edit skills and agents straight from the UI.
• Plans & tasks — plan-mode plans (filterable, searchable, editable) and the tasks Claude creates during sessions, grouped per session.

• Embedded terminal — pick up any session in a real terminal inside the app, side by side with its read-only Lens view.
• In-app chat — continue a conversation through the Claude Agent SDK without leaving ClaudeLens: streaming replies, native slash commands, and interactive tool approvals (Allow / Always / Deny). Chats persist to the same transcript files as the terminal, so the two are interchangeable.

• Live Monitor (experimental) — real-time view of active Claude processes: status (idle / thinking / busy), a sliding activity chart, tool-frequency breakdown, and elapsed timer.
• Duplicate projects — detect project folders that point to the same directory and merge their history and memory.

Everything updates live: any change under ~/.claude/ while you work in the terminal is reflected immediately, and vice versa.

• macOS 12 Monterey or later, or Linux / Windows (experimental)
• Claude Code installed and used at least once (so ~/.claude/ exists)

macOS — download the .dmg from the Releases page, open it, and drag ClaudeLens to Applications.

Linux (experimental) — download the .AppImage from the Releases page and make it executable with chmod +x. Opening a session in a terminal relies on a common terminal emulator being installed (gnome-terminal, konsole, xfce4-terminal, or xterm).

Windows (experimental) — download and run the .exe installer from the Releases page. Opening a session launches it in a new cmd window.

First launch — Gatekeeper warning

The app is not code-signed. macOS will block it on first open. Right-click the app in Finder and choose Open, then confirm in the dialog.

Alternatively, run this command once in Terminal:

xattr -d com.apple.quarantine /Applications/ClaudeLens.app

npm install
npm run dev              # Vite dev server + Electron
npm run electron:build   # Generate distributable DMG

• The embedded terminal and in-app chat require the claude CLI installed and in PATH
• Session list is not paginated — may be slow with very large histories (500+ sessions)
• No automatic updates
• App is not code-signed (see installation note above)
• On Windows, the Live Monitor and background agents are not yet supported (they rely on Unix process tooling); browsing sessions, memory, CLAUDE.md, and opening sessions in a terminal all work

MIT