Show HN: Memnest, local-first memory shared by pi, Claude Code and Codex

A developer going by blue-b posted Memnest to Hacker News on August 30, 2026, as a Show HN submission — a local-first memory tool designed to be shared across Raspberry Pi, Claude Code, and OpenAI Codex, per the HN listing.

The project, hosted at GitHub, positions itself as a shared memory layer that multiple AI coding agents and edge devices can read from and write to without relying on a remote server. The "local-first" framing indicates that state is stored on-device, with the goal of keeping context persistent across different tools in a developer's workflow — whether that's a Raspberry Pi running an automated task, a Claude Code session, or an OpenAI Codex integration.

The source listing provides no information about the author's background, team size, prior funding, or commercial status. The project appears to be an open-source, solo or small-team effort based on the single GitHub repository and the informal Show HN format.

The problem Memnest targets is a practical one in AI-assisted development: each coding agent typically maintains its own isolated context, so switching between tools means losing accumulated project knowledge. A shared, local memory store would allow, for example, a Raspberry Pi running an agentic loop to surface the same project state that a Claude Code session last wrote — without any data leaving the local network.

The HN post had 1 point and no comments at the time of submission, suggesting it is very early in public visibility. No roadmap, documentation details, or future plans are mentioned in the available source material beyond what is captured in the repository title and HN headline.