MetaBot CLI
The MetaBot CLI connects local coding agents to messaging and a web console while retaining tasks and sessions
- 1,003 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Bots execute commands and edit files under the selected engine’s permissions
- Conversations and results pass through messaging platforms and the model
- A compromised channel or console token can expose agent access
Install
- Works in
Manual install
curl -fsSL https://github.com/xvirobotics/metabot/releases/latest/download/install.sh | bashInstalls the full personal environment. Review the script and background services before running.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Bots execute commands and edit files under the selected engine’s permissions
- Conversations and results pass through messaging platforms and the model
- A compromised channel or console token can expose agent access
This is third-party code. Review the repository files before installing.
What it does
MetaBot receives tasks from a messaging app and passes them to a coding agent on your machine. For each bot, you select an agent CLI and working directory, then supply channel credentials. One installation can run several bots for different projects. MetaBot starts Codex CLI by default. Kimi Code integration and Claude Code compatibility are also available. The web console shows responses, tool calls and generated files; users can answer agent questions or stop a task there. The CLI manages MetaBot services and sends messages to agents. Additional features organize agents into teams with shared memory and long-running tasks. Telegram, Lark or Slack provide the user-facing messaging channel.
Who it is for. For developers who want to assign work to their own coding agents through chat and follow progress from a phone.
Good fit when
- A long task needs to start in a messaging app and continue in a local console
- Different projects need separate bots and working directories
- Several agents need task exchange and shared memory across sessions
- You want tool calls and their results in one task history
Not a fit when
- You need a hosted agent without keeping your own machine running
- Other people can access the chat while the bot can modify a work repository
- You need to steer Codex through Lark while a task is running: that adapter capability is still in development
Example request
Connect one bot to a project copy and show how it receives a task, asks a question and returns a result in the console.Limitations
Full installation needs Node.js 22.19 or later, Git and a coding agent with its own sign-in and model access. The installer adds Core, a chat bridge and background services that run separately from the conversation window. The messaging platform and model provider receive their parts of the dialogue even when the agent runs on your computer. Slack requires a public HTTPS event endpoint; without one, the channel cannot deliver tasks to the installed bridge.
How to disable. Stop MetaBot services using the installation guide and revoke connected bot tokens. Preserve configuration and data under `~/.metabot/` and `~/.metabot-core/` before removal. Manage coding-engine authentication separately.
FAQ
How do I install MetaBot?
Use the installation script below after reviewing it. Run `metabot status` and `metabot health` to inspect the prepared environment.
Where is the local console?
It defaults to `http://localhost:9200`. Sign-in requires the token in `~/.metabot-core/token`; keep it out of shared chat.
How do I choose engines for different bots?
Use `bots.json`. Codex is selected when an engine is omitted. Each bot retains separate channel, workspace and session settings.
Does Telegram require a public server?
No. Telegram commands use long polling, without a public inbound event endpoint. Keep the bridge machine running and connected to Telegram.
How do I update the installation?
Use `metabot update` for a packaged installation or `metabot update --git` for a source checkout. GitHub Release updates preserve `.env` and `bots.json`. The package-managed `~/.metabot/default.env` may change with updated defaults.
Does MetaBot include model usage?
No. Agent subscriptions and API usage are separate; MetaBot code uses MIT.
Discussion
No questions or reviews yet.
Related
CLI toolsbotmux connects Lark and Feishu to Claude Code, Codex, Gemini CLI and OpenCode: each message starts its own agent session with live cards
An AI-agent workspace with parallel worktrees, cross-provider orchestration, and bots across chat channels.
GitBot runs Claude Code, Codex or OpenCode locally as reusable bots with persistent threads and permissions you control.
A bridge linking WeChat, WeCom or Telegram with local Codex, Claude Code, OpenCode and Pi so the agent can be driven from a phone.
A bridge between local CLI agents: Codex, Cursor or Claude Code delegate work to Antigravity, Grok, Kimi and seven more.
Codync turns Claude Code, Codex, Cursor and other agents on your computer into bots you message from an iPhone or a terminal
README in short
The Core server stores its data separately from the messaging bridge. The installer places bundled skills in the selected coding agent’s directories.