What is op7418/Claude-to-IM-skill?

Claude-to-IM Skill by op7418 is an MIT-licensed skill that runs a background daemon bridging Claude Code or Codex to five instant-messaging platforms. Messages from your chat app reach the coding agent, and its responses, tool use and permission prompts come back, so you can drive a coding agent from your phone.

Can I use op7418/Claude-to-IM-skill commercially?

Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.

Is op7418/Claude-to-IM-skill still maintained?

Yes. The repository last received commits 176 days ago.

What programming language is op7418/Claude-to-IM-skill written in?

Mainly TypeScript, according to GitHub's language statistics.

Claude-to-IM Skill: Claude Code on Telegram, Discord, WeChat

At a glance

What is it?: Claude-to-IM Skill by op7418 is an MIT-licensed skill that runs a background daemon bridging Claude Code or Codex to five instant-messaging platforms. Messages from your chat app reach the coding agent, and its responses, tool use and permission prompts come back, so you can drive a coding agent from your phone.
Who is it for?: Install Claude-to-IM Skill if you want to reach Claude Code or Codex from a chat app rather than a terminal. Add it with npx skills, run the setup wizard to enter bot tokens, and start the daemon; approve tool calls from inline buttons or /perm replies.
Can I use it commercially?: Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?: Yes. The repository last received commits 176 days ago.
What is it written in?: Mainly TypeScript, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.

DEEP OPEN-SOURCE ANALYSIS

Hysen Labs Editorial · September 15, 2026 · op7418/Claude-to-IM-skill

What it does and who it is for

Claude-to-IM Skill bridges Claude Code or Codex to instant-messaging platforms, so you can chat with an AI coding agent from Telegram, Discord, Feishu/Lark, QQ or WeChat. It runs a background daemon that connects your IM bots to Claude Code or Codex sessions: messages from chat are forwarded to the agent, and responses, including tool use, permission requests and streaming previews, are sent back to your chat.

The README places it deliberately as the lightweight option. It was extracted from CodePilot, a full desktop GUI app by the same author, for users who prefer a CLI-only setup. So the audience is people who already run Claude Code or Codex and want to reach it from their phone or a team chat, kick off a task, approve or deny its actions, and read results, without sitting at the terminal. It is a skill, not a standalone program: it installs into Claude Code or Codex and is driven from there.

How the bridge works

The architecture is a simple chain, and the README draws it plainly: you in an IM app, talking over the platform's Bot API to a Node.js background daemon, which talks through the Claude Agent SDK or the Codex SDK to Claude Code or Codex, which reads and writes your codebase. Which runtime it uses is configurable through the CTI_RUNTIME variable, with claude, codex or auto.

That design has one consequence worth understanding before installing: the bridge hands a chat conversation real control over your machine's codebase. The README addresses this with a permission model rather than by removing the risk. Tool calls require explicit approval, through inline Allow/Deny buttons on Telegram and Discord, or text /perm commands and quick 1/2/3 replies on Feishu, QQ and WeChat. Conversations persist across daemon restarts, and secrets are protected: tokens are stored with chmod 600 and auto-redacted in all logs.

Installing it in Claude Code or Codex

Installation differs by which agent you use. For Claude Code the recommended route is npx skills:

bash

npx skills add op7418/Claude-to-IM-skill

Then you tell Claude Code /claude-to-im setup. You can instead clone straight into the skills directory (git clone into ~/.claude/skills/claude-to-im), which Claude Code discovers automatically, or symlink a development checkout.

For Codex the recommended route is the install script, which places the skill under ~/.codex/skills/claude-to-im, installs dependencies and builds the daemon:

bash

git clone https://github.com/op7418/Claude-to-IM-skill.git ~/code/Claude-to-IM-skill
bash ~/code/Claude-to-IM-skill/scripts/install-codex.sh

After that you tell Codex claude-to-im setup. To verify, in Claude Code you start a session and type / to see claude-to-im in the skill list; in Codex you say claude-to-im setup or start bridge. The prerequisites are Node.js 20 or newer, plus whichever CLI you target: an authenticated Claude Code (the claude command available), or Codex installed with npm install -g @openai/codex and authenticated with codex auth login.

Setup, start and chat

The daily flow is three steps. Setup runs a wizard that guides you through choosing channels (any combination of the five platforms), entering credentials with inline instructions on where to get each token and which settings to enable, setting defaults for working directory, model and mode, and validating tokens against the platform APIs immediately. Then you start the daemon:

text

/claude-to-im start

In Codex the natural-language equivalent is start bridge. The daemon runs in the background and survives closing the terminal. From there you open your IM app, message the bot, and Claude Code or Codex responds through the bridge; when it wants to edit a file or run a command, the permission prompt appears in the chat.

A small set of management commands rounds it out, each available as a Claude Code slash command and a Codex natural-language phrase: stop, status, logs (with an optional line count), reconfigure and doctor for diagnosing issues. The doctor command is the first thing to reach for if the bridge does not respond.

Platform setup and the permission model in practice

The setup wizard carries inline guidance for every platform, which matters because each IM has its own bot-creation dance. For Telegram, you message @BotFather, run /newbot, copy the token, optionally disable privacy for group use, and find your user ID via @userinfobot. For Discord, you create an application in the Developer Portal, reset and copy the bot token, and enable the Message Content Intent. The other platforms follow the same pattern of creating a bot and pasting its credentials.

The permission model is the feature to configure with care, because it is the safety boundary. On Telegram and Discord you get real Allow/Deny buttons; on Feishu, QQ and WeChat you approve with a /perm command or a quick numbered reply. Streaming preview, seeing Claude type its response, is available on Telegram and Discord. Treat every permission prompt as the moment where a chat message becomes a command on your machine, and do not run the bridge in a chat that untrusted people can post to.

Maturity, licence and cautions

The project is MIT licensed and early in its life: package.json marks it version 0.1.0, and no releases have been tagged. It ships an evals/ directory and a test suite, and depends on the Anthropic Claude Agent SDK with the OpenAI Codex SDK as an optional dependency, which matches the configurable-runtime design. The last push was on 2026-03-23.

The cautions follow from what the tool is rather than from any flaw in it. Bridging a coding agent to a messaging platform means bot tokens that, if leaked, let someone reach your agent, and a permission prompt that, if approved carelessly from a phone, runs real commands on your codebase. The README's own safeguards, chmod 600 tokens, log redaction and mandatory approval, are the right ones, and the sensible way to use the skill is to keep the bot private, review each permission request, and start with a scoped working directory rather than your whole machine.

Editorial conclusion

Install Claude-to-IM Skill if you want to reach Claude Code or Codex from a chat app rather than a terminal. Add it with npx skills, run the setup wizard to enter bot tokens, and start the daemon; approve tool calls from inline buttons or /perm replies. Keep in mind it gives a chat bot control over your codebase, so guard the bot tokens and the permission prompts carefully.

Frequently asked questions

What is the Claude-to-IM Skill?

It is a skill for Claude Code or Codex that runs a background daemon bridging the agent to Telegram, Discord, Feishu/Lark, QQ or WeChat. You chat with your coding agent from the IM app, and its responses, tool use and permission prompts come back into the chat.

How do I install and start it?

In Claude Code, run npx skills add op7418/Claude-to-IM-skill, then /claude-to-im setup and /claude-to-im start. In Codex, run scripts/install-codex.sh, then say claude-to-im setup and start bridge. It needs Node.js 20+ and an authenticated Claude Code or Codex CLI.

Is it safe to give a chat bot access to my code?

The bridge enforces explicit approval for every tool call, via Allow/Deny buttons or /perm replies, stores tokens with chmod 600, and redacts them in logs. The risk is real, so keep the bot private and review each permission prompt before allowing it.

Official sources

Community notes