Claude Code ↔ Codex handoffs

Switch AI coding assistants without starting over.

loghop captures local assistant sessions, redacts sensitive context, builds a shared project timeline, and writes handoff packets so the next run resumes cleanly.

PyPI package version Supported Python versions CI status MIT license
$ loghop run --provider claude
capturing session S-001...
 timeline updated
 handoff H-001 created

$ loghop run --provider codex
loading previous context...
 resumed from S-001

Why loghop?

AI coding tools are useful in different moments. loghop keeps project context portable so changing assistants does not mean rebuilding the conversation from memory.

Capture

Reads native Claude Code and Codex transcripts and stores clean local session records.

Redact

Removes sensitive context before session summaries, timelines, or handoffs are reused.

Resume

Turns previous work into actionable context for the next assistant run.

The workflow

Run an assistant

Work normally with Claude Code or Codex inside your repository.

Capture the session

loghop records status, summary, decisions, todos, changed files, and transcript context.

Build a handoff

The project timeline becomes a compact packet for the next run.

Resume elsewhere

Start another assistant with the relevant context already prepared.

Install and start

loghop is published on PyPI and works with Python 3.12 or newer.

$ pipx install loghop
$ cd your-repo
$ loghop init
$ loghop run

Designed for real projects

Local-first storage

Project state lives in .loghop/. Handoffs, timelines, sessions, and redacted transcripts stay under your control.

CLI and TUI

Use compact commands for automation or the Textual terminal UI for browsing sessions and handoffs.

Provider-aware

Capture Claude Code and Codex sessions while keeping a shared project memory across both tools.

Release-ready

CI, security checks, packaging smoke tests, and reproducible handoff files are built into the project workflow.

Documentation

The docs follow the Diátaxis model: start with a tutorial, use how-to guides for tasks, reference for exact behavior, and explanations for design decisions.