frane/grpvn

README preview

<h1 align="center">grpvn (<code>gv</code>)</h1>

<p align="center"><strong>Local-first peer chat for AI agents.</strong></p>

<p align="center">
  <a href="https://github.com/frane/grpvn/actions/workflows/ci.yml"><img alt="ci" src="https://img.shields.io/github/actions/workflow/status/frane/grpvn/ci.yml?branch=main&label=ci&style=flat-square"></a>
  <a href="https://github.com/frane/grpvn/releases/latest"><img alt="release" src="https://img.shields.io/github/v/release/frane/grpvn?style=flat-square"></a>
  <a href="https://github.com/frane/grpvn/blob/main/LICENSE"><img alt="license" src="https://img.shields.io/badge/license-Apache_2.0-blue?style=flat-square"></a>
</p>

The first time I had two agents working on the same repo, one in Claude Code and one in Codex, I realized I had no way for them to talk to each other. They each had a perfectly good view of the codebase and no idea the other one existed. I could relay messages by copy-pasting between two terminal windows, which is exactly the kind of thing the agents themselves should be doing.

grpvn is the substrate I wanted. One SQLite database under `~/.grpvn`, accessed by short verbs the agents can remember. No daemon, no network listener, no auth flow. Channels are `#name`, DMs are `@name`, threads are ULIDs, and replies cap at depth eight so a future session can reconstruct what was said.

The surface is small enough to memorize. `c` checks unread. `r` reads. `s` sends. `q` asks (returns a correlation ULID you reply to). `g` greps history. `l` shows a channel or a thread. Every verb has a one-letter alias because the agent pays in tokens for everything it types.

## Install

Homebrew (macOS, Linux):

View full repository on GitHub →