Claude Code Sessions History

Browse, search and export every Claude Code session on your machine, from inside VS Code.

Claude Code writes each session to a JSON Lines transcript under ~/.claude/projects and gives you no way to read it back. This extension does: a docked tree of your sessions, a full browser panel with the conversation, deep search across every message, statistics, Markdown export, and one click to resume a session in a terminal.

Everything runs locally. Nothing is uploaded, there is no telemetry, no account, and no paid tier — every feature listed here is in the extension you install.

Features

Two ways in. A Sessions tree in the activity bar for navigation while you work, and a full browser panel (Ctrl+Shift+H / Cmd+Shift+H) for reading, searching and statistics.

Real session titles. Claude Code generates a title for each session; sessions that never got one fall back to your first message. Sessions group by date or by project, whichever you prefer.

The whole conversation. Your messages and Claude's, collapsible thinking blocks, every tool call with its input and output, and inline diffs for every file Edit and Write — read from the structured patch Claude Code records, not reconstructed from text.

Files a session produced. Artifacts and Markdown documents appear as cards: open the published artifact in your browser, preview the file, or open it in the editor.

Filtering inside a session. A session of several thousand messages has a filter box of its own: type to keep only the messages that match, across message text, thinking, tool calls, diffs and the files a session produced, with every match highlighted. Terms combine in any order, quotes match a phrase, and you can narrow to what you said, what Claude said, tool calls, notifications, or only the messages that changed a file. Claude Code injects notices as user-role messages, so From you leaves those out and keeps to what you actually typed.

Search that reaches the message bodies. Filter by title, project or branch instantly, or turn on deep search to scan the text of every message in every transcript. Results stream in as they are found and clicking one jumps to that message.

Statistics. Sessions per day, model and branch distribution, busiest projects, tokens processed, and total transcript size.

Export to Markdown. One session or a date range, with thinking and tool calls folded into <details> blocks and file edits kept as diffs.

Resume. Continue a session in a Claude Code tab, using the Claude Code for VS Code extension, when the session belongs to the folder open in this window. Any other session is reopened in a terminal with claude --resume, started in that session's own project directory so Claude Code can actually find it. Where to resume, and the terminal command, are configurable.

Fast on real histories. Session metadata is cached against each transcript's size and modification time, so only changed transcripts are re-read. Conversations are paged into the view rather than rendered whole — a 60 MB transcript opens without freezing the panel — and Load everything renders the rest of a long session in one go.

Usage

1. Install the extension.
2. Press Ctrl+Shift+H (Cmd+Shift+H on macOS), or open the Sessions view from the activity bar.
3. Pick a session to read it.

Everything is also available from the command palette under Claude Sessions.

Requirements

• Claude Code installed and run at least once
• VS Code 1.85.0 or later

Settings

Privacy

The extension reads files under your Claude Code home directory and writes nothing back to it. It makes no network requests of any kind, collects no telemetry, and needs no sign-in. The only data it stores is a metadata cache and your bookmarks, both in VS Code's own extension storage.

Two actions reach outside VS Code, and only when you ask for them: Open in browser on an artifact card opens a claude.ai link, and Resume continues the session in the Claude Code extension or a terminal running Claude Code.

How it works

Claude Code appends one JSON object per line to <claude home>/projects/<encoded project path>/<session id>.jsonl. Reading those back has two wrinkles worth knowing about, because they are where naive transcript readers go wrong:

The encoded folder name is not a path. Claude Code replaces every path separator with -, but it replaces _ and . the same way, so C:\Projects\App_2025 and C:\Projects\App\2025 both encode to C--Projects-App-2025. Decoding is therefore guesswork. This extension reads the cwd field recorded inside the transcript instead, which is exact, and only falls back to decoding the folder name — verified against the filesystem — when no record carries one.

Not every user record is from you. Claude Code stores tool results, slash-command plumbing, IDE state and injected reminders as user-role messages. Showing those as things you said makes a transcript unreadable, so they are filtered out. A prompt you typed while Claude was still working is stored differently again, as a queued attachment; those are kept and marked.

Development

npm install
npm run compile      # or: npm run watch
npm test             # parser tests, plain Node, no Extension Host needed
npm run icon         # re-render media/icon.png from the brand values
npm run package      # build a .vsix

Press F5 in VS Code to launch an Extension Development Host.

The code is laid out as src/claude (reading transcripts), src/store (indexing, caching, search) and src/ui (tree, panel, export), with the webview front end in media/.

Acknowledgements

The idea of a Claude Code history browser for VS Code is not original to this extension — Claude Code History by doorsofperception got there first. This is an independent implementation, written against the current Claude Code transcript format and released with every feature unlocked.

License

MIT. See LICENSE.txt.

Built by Methodox Technologies, Inc.