Multiplexes messaging channels to AI agents through a single CLI and a file-based job queue. Nothing else required.
ingress egress
┌──────────────┐ ┌───────────────┐
│ channel → │ │ → channel │
│ save message │ nq queue │ watch for │
│ + attachments│──────────→────│ completed │
│ enqueue job │ │ jobs, reply │
└──────────────┘ └───────────────┘
↕
nq → dispatch
(reads prompt.txt,
runs coding agent)
Requires Deno (v2+), nq, and a coding agent (Pi).
Create a Telegram bot
and grab the bot token, then create ~/.config/muxclaw/config.json:
{
"channels": {
"telegram": {
"token": "bot123:ABC..."
}
},
"allowedUsers": [
{ "userId": "12345" }
],
"workspace": "/path/to/your/project",
"agent": {
"name": "pi",
"stream": true
}
}Then, in two terminals:
deno task cli ingress # receive messages, enqueue jobs
deno task cli egress # watch for results, send repliesdocker run -it --rm \
-v ~/.config/muxclaw:/home/deno/.config/muxclaw \
-v $(pwd)/workspace:/workspace \
ghcr.io/jihchi/muxclawLicensed under either of
at your option.