search-conversations

search-conversations

A Claude Code skill that searches your past conversations by keyword and helps you jump back into any matched session.

Every Claude Code chat is stored locally as JSONL in ~/.claude/projects/. This skill searches across all of them, shows you matching snippets with context, and tells you exactly which session to /resume.

Install

As a Claude plugin (recommended)

Inside Claude Code, run these two commands in order:

/plugin marketplace add Lohomi15/claude-search-conversations
/plugin install search-conversations@claude-search-conversations

The first command registers this repo as a plugin source; the second installs the skill from it.

Manual install

git clone https://github.com/Lohomi15/claude-search-conversations ~/.claude-search-conversations
mkdir -p ~/.claude/skills
cp -r ~/.claude-search-conversations/skills/search-conversations ~/.claude/skills/

Usage

Just ask Claude Code naturally:

• "Where did we talk about the Postgres migration?"
• "Find the chat about the Coorg trip"
• "Which session did we design the auth flow in?"

Claude will invoke the skill, show matching conversations, and offer to resume the most relevant one.

You can also run the search script directly:

python3 ~/.claude/skills/search-conversations/search.py "postgres migration"
python3 ~/.claude/skills/search-conversations/search.py "auth flow" -c 2 -n 10
python3 ~/.claude/skills/search-conversations/search.py "deploy" --since 2026-04-01

Flags:

• -c N — include N turns of context before/after each hit (default: 1)
• -n N — cap number of hits (default: 20)
• --chars N — truncate each snippet line to N chars (default: 400)
• --since YYYY-MM-DD — only include matches on or after this date
• --until YYYY-MM-DD — only include matches on or before this date

Sample output

Found 3 match(es) for 'postgres migration':

── 2026-04-18 14:22 UTC | session 7a3f19cc | -Users-you-project
  [user] can you help me plan the postgres migration from rds to supabase
  [assistant] Sure — let's start by inventorying the tables that hold user data...
  → /Users/you/.claude/projects/-Users-you-project/7a3f19cc-....jsonl

── 2026-04-17 09:05 UTC | session 2b88a041 | -Users-you-project
  [assistant] The postgres migration plan we wrote yesterday is still the plan of record
  [user] great, let's run it
  → /Users/you/.claude/projects/-Users-you-project/2b88a041-....jsonl

To jump back into a matched session, copy its full UUID and run:

/resume 7a3f19cc-....

Updating

If you installed via the plugin commands

Inside Claude Code:

/plugin marketplace update Lohomi15/claude-search-conversations
/reload-plugins

The first command fetches the latest version from GitHub; the second reloads it into your current session.

If you installed manually (git clone)

From wherever you cloned the repo:

cd <path-to-cloned-repo>
git pull

If you symlinked the skill into ~/.claude/skills/search-conversations (recommended for contributors), git pull is all you need — the symlink will pick up the new files automatically.

If you copied the files in, re-copy them:

cp -r skills/search-conversations ~/.claude/skills/

Or just run the helper:

./update.sh

How it works

1. Globs ~/.claude/projects/*/*.jsonl — every Claude Code session you've ever had.
2. Extracts text from user and assistant turns (including tool use / tool result bodies).
3. Does a case-insensitive substring match.
4. Returns ranked snippets with session IDs, timestamps, and file paths.

It's all local. No network calls, no external dependencies.

Requirements

• Claude Code
• Python 3.8+
• macOS or Linux

Issues & suggestions

Found a bug, have a feature idea, or want to suggest an improvement? Open an issue:

👉 github.com/Lohomi15/claude-search-conversations/issues

There are templates for bug reports and feature requests to make it quick. Pull requests are also welcome — fork the repo, make your change, and open a PR.

Planned improvements and known candidates are tracked in CHANGELOG.md under [Unreleased].

License

MIT — see LICENSE.