claude-code

Show HN: Agent-hop – a Rust TUI that hops a live Claude Code chat into Codex

Claude News

claude-code

Agent-hop, a terminal interface written in Rust, moves an in-progress coding session out of Claude Code and into Codex, Pi, Grok Build or OpenCode without starting the conversation over. The tool was presented on Hacker News as a Show HN, and its code is published on GitHub.

At a glance

  • The author's stated problem is context loss: switching from Claude Code to another CLI normally means abandoning the chat, and Agent-hop is built to carry that chat across instead.
  • According to its GitHub page, the repository was created on 2026-08-07 and listed 11 stars and 3 forks, with top contributor hetpatel-11 credited with 74 contributions.
  • Four alternative agents are named as hop targets, Codex, Pi, Grok Build and OpenCode, which makes agent choice something that can be changed partway through a task rather than at the start.

Agent switching is in practice a one-way door: work started with one CLI tends to stay there because the transcript does not travel with it. A tool that makes the choice reversible mid-task would change how those agents are evaluated, since lock-in through conversation history weighs more than feature parity on long sessions. On an 11-star repository, that reads as a stated goal rather than a demonstrated one.

The project did not start as a switcher. Its author says it began as a resume and search tool, and that friction in that workflow prompted a purpose-built TUI. According to the repository description on GitHub, searching and resuming local coding sessions across several agents remains part of what it does.

The Show HN post names a single comparison point, Herdr, and draws the line at live switching: Agent-hop's author states that Herdr does not allow hopping between agents mid-chat. The post offers no feature-by-feature comparison and names no other tool in the category.

The repository is new by any measure. According to GitHub, it was created on 2026-08-07 and listed 11 stars and 3 forks at the time of checking, with hetpatel-11 as the top contributor at 74 contributions. Nothing in the material points to a tagged release.

How the handoff works is unstated The Show HN post describes the outcome, hopping mid-chat, but not the mechanism: whether transcripts are translated between agent formats, how much of the context survives a hop, and what happens to tool state are all left open. No version, license or roadmap is mentioned, and no date is given for when the list of supported agents might grow beyond the five named.

Comments

No comments yet. Be the first.

Join the conversation

Sign in with Google to leave a comment. Your name and avatar come from your Google profile, and the comment appears after moderation.

We only use your name and avatar from Google. We never store your email address.