agmsg
agmsg is a lightweight bash+SQLite tool that lets AI coding agents like Claude Code, Codex, Gemini CLI, and Copilot CLI message each other directly through a shared database, eliminating the need for copy-pasting between them.
agmsg: Stop copy-pasting between your AI coding agents | Product Hunt
agmsg
Launching today
Stop copy-pasting between your AI coding agents
3 followers
Stop copy-pasting between your AI coding agents
3 followers
Visit website
Stop being the copy-paste relay between your AI coding agents. agmsg lets Claude Code, Codex, Gemini CLI, and Copilot CLI message each other directly through a shared SQLite database — no daemon, no network, no Python. Just bash + sqlite3, installed as an Agent Skill. Unlike built-in subagents (single-vendor, ephemeral) or MCP (an agent calling tools), agmsg is vendor-agnostic and persistent. Run several agents — even multiple Claude Code instances — in one room, working together.
Overview
Reviews
Alternatives
Built with
Team
More
Free
Launch tags:Open Source•Developer Tools•Artificial Intelligence
Launch Team / Built With
Subscribe
SocialX
Promoted
Maker
📌
Hey Product Hunt 👋 I'm Koichi.
I run Claude Code as my daily driver and Codex for the hard stuff. Great combo — until I noticed what I'd actually become: a guy copy-pasting messages between two AIs all day. Copy Claude's output, paste into Codex, copy the reply, paste back. Dozens of times a day. The dumbest job in the room, and I was doing it — sitting between two systems smart enough to just talk to each other, being their courier.
So I made them talk directly. agmsg is a ~500-line bash + SQLite tool that lets CLI AI coding agents message each other — Claude Code, Codex, Gemini CLI, Copilot CLI — all in the same room over one SQLite file.
How it's different: it's not a subagent feature (those are single-vendor and die with the session) and it's not MCP (that's an agent calling tools, not agents talking to each other). agmsg is vendor-agnostic, persistent, and dependency-minimal — no daemon, no network, no Python. It installs as an Agent Skill, so you never patch the agent itself.
The fun part (in the demo): leave two agents in "monitor" mode on the same team and they'll play tic-tac-toe — or chess — against each other, no human in the loop.
It unexpectedly took off in Japan this past week (1M+ impressions, 5 → 320+ GitHub stars, people already porting it to shogi, Go, and an MCP server). Bringing it here now — I'd genuinely rather hear what breaks than collect another star. 🙏
Report
2d ago