Show HN: Ctxbin – A deterministic CLI for reliable AI agent handoffs

Posted by superlucky84 on Saturday January 31, 2026 @01:42PM
from the show-hn dept.

Hi HN,

I built ctxbin, a minimal CLI that lets AI agents reliably save and load shared context using branch-scoped keys inferred from git.

The problem I kept hitting was that handoffs between AI agents (or even between sessions) were fragile and non-repeatable. ctxbin solves this by storing structured context, agents, and skills in Upstash Redis with deterministic keys.

Key ideas: - Branch-scoped context ({repo}/{branch} inferred automatically) - Explicit save/load semantics (no hidden state) - Reusable agents and skills (string, directory bundles, or GitHub refs) - Designed for agent-to-agent handoff, not humans

It works well with tools like Claude Code, Codex CLI, or any AI agent workflow. Feedback very welcome.

Docs: https://superlucky84.github.io/ctxbin/

( 1 comment | on HN )

Re:Show HN: Ctxbin – A deterministic CLI for reliable AI age... by superlucky84 on Saturday January 31, 2026 @01:43PM (#46839395)

Built this after repeatedly losing context between multiple AI agents working on the same repo and branch.

I wanted something boring, explicit, and repeatable — no hidden memory or session magic. Curious how others here handle AI agent handoffs or shared context today.

[ Parent | on HN ]