history

Index Claude Code session history for semantic and keyword search.

Updated Aug 27, 2026
One-click install
npx skills add https://github.com/robdmc/claude_tools --skill history-robdmc
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: history
Source: https://github.com/robdmc/claude_tools/tree/main/history/skills
Command: npx skills add https://github.com/robdmc/claude_tools --skill history-robdmc

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires lancedb, pyarrow, sentence-transformers, numpy, and includes scripts (resource) components.

What problem does it solve?

This Skill enables users to quickly locate and analyze Claude Code conversation history across multiple projects, reducing time spent digging through sessions.

Core Features & Use Cases

  • Hybrid search: Combines semantic understanding with keyword search to surface relevant sessions.
  • Progressive disclosure: Presents results with counts, summaries, and drill-down capabilities to minimize token usage.
  • Export/Import: Export sessions to Markdown/JSON and import sessions for /resume to continue conversations.
  • Session exploration: Inspect session details, prompts, and tool usage to understand context.

Quick Start

  • Invoke: /history <query>
  • List recent sessions: /history list
  • Export a session: /history export <session_id>

Frequently Asked Questions about history

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I search Claude Code conversation history across multiple projects?

You can search Claude Code conversation history across projects using hybrid semantic and keyword search. The Skill indexes session data with LanceDB and sentence-transformer embeddings, surfacing relevant sessions with summaries and drill-down capabilities.

Can I export Claude Code sessions to Markdown or JSON for /resume?

Yes, you can export Claude Code sessions to Markdown or JSON formats. The Skill also supports importing these exported sessions, allowing you to use the /resume command to continue previous conversations.

Do I need LanceDB and Python to index Claude Code session history?

Yes, indexing Claude Code session history requires local Python tools, LanceDB, pyarrow, numpy, and sentence-transformers. These dependencies handle the local vector database storage and generate the semantic embeddings for search.

What's the best way to find a specific prompt within past Claude Code sessions?

The best way to find a specific prompt is using hybrid search, which combines semantic understanding with keyword matching. It provides progressive disclosure with counts and summaries to inspect session details while minimizing token usage.

Does semantic search on Claude Code history work offline?

Semantic search on Claude Code history works locally using LanceDB and sentence-transformers. The indexing and embedding processes run on your machine, requiring the specified Python dependencies to function.

Why does searching Claude Code history return too many tokens?

Searching Claude Code history returns too many tokens when not using progressive disclosure. This Skill minimizes token usage by presenting results with counts, summaries, and drill-down capabilities instead of loading full session transcripts.