jupyterlab-claude-code-extension 1.0.26

Browse, resume, and manage your Claude Code CLI sessions from a JupyterLab side panel. One click reactivates the right terminal - no duplicate tabs, live remote-control indicator, and favourites for the projects you keep coming back to.

jupyterlab_claude_code_extension

Manage Claude Code CLI sessions from inside JupyterLab. A left-sidebar panel lists every project under ~/.claude/projects/ deduplicated to one row per folder, marks live remote-control sessions with a green dot, and lets you jump back into any session by opening (or reactivating) a terminal pwd'd to that project and auto-running claude --resume <id>.

Features

• Three-section side panel - Favourites, Recent (top 10 by activity), and All. Each section scrolls independently; Favourites disappears when empty
• Live remote-control indicator - green dot on rows whose ~/.claude/sessions/<pid>.json is alive (verified via os.kill(pid, 0))
• One-click resume - click a row to find an existing terminal pwd'd to that project (queried server-side from the pty's process tree) and reactivate its tab; only spawns a fresh terminal if none matches. Concurrent rapid clicks are coalesced
• Smart name resolution - shows the user-set /rename name when available; auto-detected names (volatile across the same sessionId or 3+ token lowercase-kebab) fall back to the folder basename. Toggle the behaviour via the resolveSessionNames setting
• Path-segment disambiguation - when two sessions share the same display name, the row reveals the minimum number of trailing path segments needed for each to be unique
• Favourites - star a session via the right-click menu; persisted server-side at ~/.claude/jupyterlab_claude_code_extension.json
• Layout restorer - panel visibility persists across JupyterLab reloads
• Auto-disabled when the claude binary is not on PATH
• Hover tooltip with relative path (vs JL root), last activity, message count, branch, first prompt, session id

Requirements

• JupyterLab >= 4.0.0
• Python >= 3.10
• claude CLI on PATH

Install

Developers must install via the project Makefile (which orchestrates clean, build, and pip install of the resulting wheel):

make install

End-users can install the published package from PyPI:

pip install jupyterlab_claude_code_extension

Uninstall

pip uninstall jupyterlab_claude_code_extension