ClaudeClaw plugin
A plugin that makes Claude Code a background assistant: scheduled tasks, Telegram, Discord and Slack replies, a web dashboard.
- 1,332 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The daemon runs commands from messengers with your user permissions, and the full-access level opens the whole system.
- Telegram, Discord and Slack bot tokens are stored in `settings.json` and give control of the agent to whoever obtains them.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/claudeclawDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The daemon runs commands from messengers with your user permissions, and the full-access level opens the whole system.
- Telegram, Discord and Slack bot tokens are stored in `settings.json` and give control of the agent to whoever obtains them.
This is third-party code. Review the repository files before installing.
What it does
ClaudeClaw runs Claude Code as a background daemon that never sleeps. Through it the agent performs scheduled tasks, replies to messages in Telegram, Discord and Slack, transcribes voice commands and works with any connected service. The author calls it a lightweight version of OpenClaw built into Claude Code. Install goes through the Claude Code plugin marketplace, after which `/claudeclaw:start` runs a setup wizard: model, periodic checks (heartbeat), messengers and security level. There are timezone-aware cron jobs, quiet hours, separate sessions for each Discord thread and a fallback to GLM models if the main limit runs out. A web dashboard shows jobs, runs and logs. The author provides four access levels from read-only to full system access. Since version 1.0.26 an empty allowed-users list blocks everyone, and since 1.1.0 dashboard API requests require a token. The author explains that the tool wraps local Claude Code and needs no third-party authorization.
Who it is for. People who want Claude Code kept running on a server or laptop and controlled from a messenger, with regular jobs that need no manual start.
Good fit when
- You need a daily report or a repository check on a schedule with no involvement from you.
- You want to send a task to the agent from Telegram and get the answer there.
- You need to take voice commands and transcribe them for the agent.
- A team talks to the agent in Discord and each topic needs its own isolated session.
- You want to see runs and logs in a web dashboard.
Not a fit when
- You need a one-off request, where plain Claude Code is enough and a daemon only adds exposure.
- The bot would be open to an unlimited group of people, while an allowed-user list is required.
- Your plan allows only interactive use: check scheduled automated runs against the terms of your subscription.
Example request
Start ClaudeClaw, connect a Telegram bot and set up a daily 9 am job that sends me a summary of repository changes.Limitations
Through messengers the agent gets commands with your permissions, so put only your own ids in `allowedUserIds`, otherwise the daemon will not start. Access levels above read-only allow writing and running commands in the system. Discord attachments are cut to 2048 bytes. The web dashboard is protected by a token from `.claude/claudeclaw/web.token`. The daemon spends your Claude Code subscription limits. The author separately asks not to use the project for illegal actions.
How to disable. Stop the daemon, run `claude plugin uninstall claudeclaw`, remove bot tokens from `settings.json` and delete the `.claude/claudeclaw/` directory.
FAQ
What can ClaudeClaw do?
Run scheduled tasks, reply in Telegram, Discord and Slack, take voice messages and show runs in a web dashboard. Everything Claude Code can do is available here.
How do I limit access?
Fill `allowedUserIds` for Telegram and Discord and pick a security level in the wizard. An empty list blocks everyone.
Does it break Anthropic's terms?
The author says no: the tool wraps local Claude Code and needs no third-party sign-in. Check your own subscription terms yourself.
What if the limit runs out?
You can turn on a fallback to GLM models so work continues.
Is there a web interface?
Yes, a panel shows jobs, runs and logs. Its API requires a token created on first start.
Discussion
No questions or reviews yet.
Related
PluginsUpgrade for the official Claude Code Telegram plugin: voice messages, live progress, buttons, memory and an agents board
A Telegram bot drives Claude Code on your computer: you message the chat and the agent edits the project and streams the reply.
A Claude Code plugin that runs tasks on a schedule: code reviews, security checks and summaries through launchd, cron or Windows Task Scheduler
Connects Claude Code, Cursor, Codex or OpenCode to Telegram, Slack and Discord: teammates message a bot, an agent replies
A Claude Code plugin that takes a to-do list from task.md and runs it on the usage window timer, leaving an action log.
An MCP channel that passes WeChat messages into a Claude Code session and sends the agent's replies back to the same chat
README in short
The author presents ClaudeClaw as a lightweight OpenClaw inside Claude Code with a five-minute install. MIT license, written in TypeScript.