agent-message

Cheap and fast messaging between AI agents. No server, no MCP, no token, no daemon.

Reference implementation of SAMP (Simple Agent Message Protocol). Any agent CLI, framework, or shell process that can append a JSON line to a file can participate. Claude Code is the first client; the protocol is vendor-neutral.

Three paths, one protocol

Path	Cost	When
Claude Code slash commands (/message-send, /message-inbox, /message-reply)	~1 Bash tool call per op	inside a Claude Code session
Shell function (msg send, msg, msg reply, …)	0 LLM tokens	from any terminal — humans, scripts, cron
Wrapper executable (~/.agent-message-cmd send …)	one shell call	from any other agent CLI / framework

All three speak the same on-disk format. Mix and match freely.

Quick start

git clone https://github.com/slima4/agent-message
cd agent-message
./install.sh

Then in two different repos / terminals:

Repo ARepo B

msg send repo_b "ping"

msg
[04-25 17:42] from=repo_a thread=2026-04-25-repo_a-ping: ping
1 new from: repo_a

msg reply "pong"

See Install for full setup, Use for all three paths, Design for the git-inspired internals, SAMP Spec for the wire-format contract, Limits for caveats.

Why not the alternatives

Existing solutions (mcp_agent_mail, Agent Teams, broker daemons) run an HTTP server, maintain SQLite, register agent identities, require tokens, burn tokens on polling hooks. agent-message gives you the 90 % at 1 % of the cost: a shared directory of append-only JSONL files, basename-as-identity, no setup per repo. Any agent that can spawn a subprocess or write a file can participate.