Experimental — please test and report issues.
IRC bridge for MulmoClaude. Connects to any IRC server (Libera.Chat, OFTC, self-hosted, etc.).
No API keys or bot registration needed — just pick a nickname and connect.
# Testing with mock server
npx @mulmobridge/mock-server &
IRC_SERVER=irc.libera.chat \
IRC_NICK=mulmo-bot \
IRC_CHANNELS=#your-channel \
MULMOCLAUDE_AUTH_TOKEN=mock-test-token \
npx @mulmobridge/irc
# With real MulmoClaude
IRC_SERVER=irc.libera.chat \
IRC_NICK=mulmo-bot \
IRC_CHANNELS=#your-channel \
npx @mulmobridge/irc- In channels: the bot responds only when mentioned (
mulmo-bot: what is 2+2?) - In private messages: the bot responds to everything
| Variable | Required | Description |
|---|---|---|
IRC_SERVER |
Yes | e.g. irc.libera.chat |
IRC_NICK |
Yes | Bot nickname |
IRC_CHANNELS |
Yes | CSV of channels (e.g. #mulmo,#test) |
IRC_PORT |
No | Default: 6697 (TLS) or 6667 (plain) |
IRC_TLS |
No | true (default) or false |
IRC_PASSWORD |
No | NickServ or server password |
MULMOCLAUDE_API_URL |
No | Default http://localhost:3001 |
MULMOCLAUDE_AUTH_TOKEN |
No | Bearer token |
IRC_BRIDGE_DEFAULT_ROLE |
No | Role id to seed new bridge sessions with (e.g. coder, general). Applied ONLY when a irc session first appears — once the user switches role via /role <id> the session's own role wins. Unknown role ids silently fall back to the server's default with a warn log. |
BRIDGE_DEFAULT_ROLE |
No | Same as above but shared across every bridge. Transport-specific IRC_BRIDGE_DEFAULT_ROLE wins when both are set. |
The MulmoClaude server regenerates a fresh bearer token on every startup and writes it to ~/mulmoclaude/.session-token. The bridge reads that file once at launch and keeps the token in memory — so if the server restarts while the bridge is running, the bridge keeps using the old token and every API call returns 401, silently.
Fix: set MULMOCLAUDE_AUTH_TOKEN to the same long random value on both the server and the bridge. The server uses it verbatim instead of regenerating, so the token survives restarts and the bridge stays authenticated.
# Server (one-time setup — same value across restarts)
MULMOCLAUDE_AUTH_TOKEN=long-random-string yarn dev
# Bridge (separate process / machine — same value)
MULMOCLAUDE_AUTH_TOKEN=long-random-string \
<bridge-specific-envs> \
npx <this-package>@latestRecommended: at least 32 characters of random data (the server logs a warning at startup for shorter values).
Part of the @mulmobridge/* package family.
Shared libraries:
@mulmobridge/client— socket.io client library used by every bridge below@mulmobridge/protocol— wire types and constants@mulmobridge/chat-service— server-side relay + session store@mulmobridge/relay— Cloudflare Workers webhook proxy@mulmobridge/mock-server— mock server for local bridge development
Bridges (one npm package per platform):
@mulmobridge/bluesky— Bluesky DMs over atproto@mulmobridge/chatwork— Chatwork (Japanese business chat)@mulmobridge/cli— interactive terminal bridge@mulmobridge/discord— Discord bot via Gateway@mulmobridge/email— IMAP poll + SMTP reply, threading preserved@mulmobridge/google-chat— Google Chat via MulmoBridge relay@mulmobridge/irc— IRC (Libera, Freenode, custom) ← this package@mulmobridge/line— LINE Messaging API via MulmoBridge relay@mulmobridge/line-works— LINE Works (enterprise LINE)@mulmobridge/mastodon— Mastodon DMs + mentions@mulmobridge/matrix— Matrix / Element@mulmobridge/mattermost— Mattermost@mulmobridge/messenger— Facebook Messenger via MulmoBridge relay@mulmobridge/nostr— Nostr NIP-04 encrypted DMs@mulmobridge/rocketchat— Rocket.Chat@mulmobridge/signal— Signal via signal-cli-rest-api@mulmobridge/slack— Slack Socket Mode@mulmobridge/teams— Microsoft Teams via Bot Framework@mulmobridge/telegram— Telegram bot@mulmobridge/twilio-sms— SMS via Twilio Programmable Messaging@mulmobridge/viber— Viber Public Account bots@mulmobridge/webhook— generic HTTP webhook bridge@mulmobridge/whatsapp— WhatsApp Cloud API via MulmoBridge relay@mulmobridge/xmpp— XMPP / Jabber@mulmobridge/zulip— Zulip
Published from the MulmoClaude monorepo by Receptron.
- MulmoClaude — an open-source AI assistant platform that runs on your own computer. Claude Code as the engine, a personal wiki for long-term memory, schema-driven collections for your data, and chat that summons the right GUI (markdown, charts, forms, spreadsheets, wikis) for each task.
- MulmoTerminal — a terminal-first cockpit for running many AI coding agents in parallel. One roster showing every session's summary and PR status, tmux-backed session persistence, git-worktree isolation, one-click PRs, and mobile push with remote reply.
- MulmoTerminal manual — setup, workflows, feature reference, configuration, mobile notifications, and alternative / local model providers. Available in English and Japanese.
MIT