GitHub Repository

Sync & search Claude Code + Codex CLI conversation history in SQLite. Zero dependencies.

4 starsPython

I built a tool to search 50K+ AI coding conversations in SQLite

Name: I built a tool to search 50K+ AI coding conversations in SQLite
Availability: InStock
Author: khaliqgant

by khaliqgant·Mar 8, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●●SolidNiche GemSolve My Problem

FTS5 search over 50K+ prompts is handy; incremental sync and zero dependencies keep it lightweight.

Strengths

•Zero-dependency Python script with incremental sync (byte-offset tracking) means fast re-runs and no bloat.
•FTS5 full-text search across three heterogeneous sources (Claude JSONL, Codex JSONL, Agent Relay API) in one DB is genuinely useful for developers who context-switch.
•Drill-down workflow (search → full prompt → session context → resume) respects developer muscle memory—integrates with Claude CLI directly.

Weaknesses

•Solves a niche problem only developers with existing Claude Code / Codex CLI workflows will encounter; limited audience.
•No UI—CLI only, and CLI search is commodity functionality (grep + jq existed first); the value is consolidation, not innovation.

Post Description

I'm moving to a new computer and realized that the history.jsonl claude and codex files contain some very useful and valuable previous prompt sessions. Also the agent-relay conversations contain some gold.

Details: - Imports history from Claude Code, Codex CLI, and Agent Relay into one SQLite DB - Incremental sync — only reads new bytes on re-run - FTS5 full-text search across all your prompts - Drill-down workflow: search → show full prompt → see session context → resume conversation - Runs as a launchd/cron job for continuous sync