pmarreck

Chatscan — Development skill for Claude Code

Development community

Search Claude Code/Gemini/Codex conversation history with semantic + lexical search.

How to install Chatscan

This entry records only its repository, not the path inside it, so there is no exact command to give. Open pmarreck/chatscan and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Chatscan does

Search Claude Code/Gemini/Codex conversation history with semantic + lexical search.

Alternatives in Development

  • Understand Anything — Graphs that teach graphs that impress 80.7k ★
  • Mempalace Search — Search your memories across the MemPalace using semantic search with wing/room filtering 58.6k ★
  • Serena — Code Intelligence Advanced code navigation and semantic search 14k ★

README

chatscan

[![Mechatron Prime CI](https://img.shields.io/endpoint?url=https%3A%2F%2Fthelio-nixos.tail66c90.ts.net%2Fbadges%2Fchatscan.json&style=for-the-badge)](https://thelio-nixos.tail66c90.ts.net/mechatron-prime/) [![Build](https://github.com/pmarreck/chatscan/actions/workflows/build.yml/badge.svg?branch=yolo)](https://github.com/pmarreck/chatscan/actions/workflows/build.yml) [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)

Search your AI coding conversation history with semantic + lexical + recency search.

Supports **Claude Code**, **Codex**, and **Gemini CLI** conversations. Indexes them into SQLite (with FTS5 + optional vector embeddings via **Ollama** or any **OpenAI-compatible** server like **oMLX**) and provides fast hybrid search with recency weighting.

Quick start

# Index your conversations (auto-detects Claude, Codex, or Gemini)
chatscan index

# Search (scoped to current project by default)
chatscan "SIMD optimization"

# Search all projects
chatscan "error handling" --all

# Search across all LLM sources
chatscan "error handling" --all-llms

# Regex search via ripgrep
chatscan --regex "indexOfIgnoreCase"

# JSON output
chatscan "config" --json

# Bounded, raw-verified evidence for an agent
chatscan recall "why was this changed" --project-exact "$PWD" --max-bytes 12000

Features

  • Multi-LLM support — indexes Claude (~/.claude/projects/), Codex (~/.codex/sessions/), and Gemini (~/.gemini/tmp/) conversations
  • Hybrid search — semantic (vector), lexical (FTS5), and recency signals fused by Reciprocal Rank Fusion (RRF), which ranks by rank-position rather than combining incomparable raw scores, so a strong match in any one signal reliably surfaces
  • Recency as a gentle tiebreaker — recent conversations get a mild boost (a low-weight RRF ranker), not enough to bury an exact-term match under recent-but-irrelevant chatter
  • Sandwich display — matched me