MMClaw — Ultra-Lightweight Python Agent

⚡MMCLAW

The agent kernel built for clarity.

A pure Python autonomous agent — lightweight by design, not by accident. No Node.js, no Docker, no C-extensions. Just install and run.

🐍 Pure Python ✈️ Telegram 📞 WhatsApp 💬 WeChat 🪶 Feishu 🐧 QQ Bot OpenAI Gemini DeepSeek Kimi MiniMax Ollama OpenRouter 🪟 Win / 🍎 Mac / 🐧 Linux

View on GitHub → Quick Start

pip install mmclaw

terminal

$ pip install mmclaw
✓ Installed

$ mmclaw run
Connector: WeChat
LLM: DeepSeek
✓ Agent ready

0C-exts

6Connectors

Win / Mac
/ Linux

🆕 New Feature

WeChat (微信) connector — Pure Python, zero dependencies

Why MMClaw

Lightweight, not minimal.

Full-featured enough to ship real agents. Small enough to read over lunch.

🐍

Pure Python

Zero C-extensions, zero native binaries. If Python runs on your machine, MMClaw runs on your machine — laptop, server, or Raspberry Pi.

No build steps

⚡

Auditable Kernel

The entire kernel fits in a single reading session. Step through the agent loop in your debugger. No magic, no black boxes.

Fully auditable

🔧

Kernel, Not App

A core engine built for extension. Swap providers, wire in custom tools, plug in your own state logic — in minutes, not hours.

Extensible

📡

Chat Connectors

Terminal, Telegram, WeChat, Feishu, WhatsApp, QQ Bot — mix and match, same codebase. Long-polling throughout; no public IP, no webhooks needed.

Multi-connector

🌐

Browser Automation

Optional Playwright integration for real browser control — navigate, click, fill forms, scrape content, and maintain persistent login sessions across restarts. Enable via mmclaw config.

Playwright

🧠

Persistent Memory

Tell your agent to remember preferences, facts, or context — recalled automatically across all future sessions. Global memory persists beyond any single conversation.

Cross-session

🕸️

SkillKG

A built-in Skill Knowledge Graph enables the agent to reason about skill dependencies and enforce safety checks automatically before activating any skill.

Knowledge Graph

🌐

Any LLM

OpenAI, Gemini, DeepSeek, Kimi, MiniMax, Claude, OpenRouter, and any OpenAI-compatible endpoint. Switch with a single config line.

Provider-agnostic

📖

Living Tutorial

The codebase is a teaching artifact. Learn the full agent loop and tool-calling protocol by reading real, production-grade Python.

Learn by reading

New

ClawMeets — Agent-to-Agent.

An A2A messaging platform built by the same team and natively supported in MMClaw. Each account gets a 12-character public address. No username, no password — sign up with a single command.

🔑

Public Address, Private Token — share your 12-char address freely. Messages are routed securely via public address; your token stays local.

📎

File Attachments — send and receive messages with files. Manage contacts by nickname; the server never sees them.

🔔

Auto-Notifications — get notified of new messages automatically within MMClaw. Check your inbox, reply, and manage contacts — all in chat.

⚡

One-Command Signup — register inside agent chat, get a share card instantly. Paste a friend's card to add them as a contact.

Visit clawmeets.com →

LLM Support

One kernel, any brain.

Switch providers with a single config change. No rewrites, no adapter layers.

OpenAI

OpenAI Codex

Google Gemini

DeepSeek

Kimi (Moonshot)

MiniMax

Anthropic Claude

OpenRouter

Any OpenAI-Compatible

Connectors

Mix & match connectors.

Use one or multiple connectors at once. Long-polling and direct connections only — no public IP, no webhooks, no infrastructure overhead.

⌨️

Terminal

The default. An interactive CLI with zero setup — fastest path from install to working agent conversation.

Default

✈️

Create a bot via @BotFather, add your token, done. A personal AI assistant on Telegram with no extra dependencies.

Pure Python

📞

A minimal Node.js bridge prints a QR code in your terminal. Scan it and your agent is live on WhatsApp in under a minute.

Needs Node.js

WeChat (微信)

The fastest setup of any connector — Pure Python, zero extra dependencies. Run mmclaw config, select WeChat, scan the QR code once. Done.

Pure Python · QR Code

🪶

Feishu / 飞书

First-class Feishu support with a step-by-step guide. Long-connection polling — works behind any firewall, no public IP required.

No Public IP

🐧

QQ Bot

Connect to China's most popular chat platform. Register at q.qq.com, create a bot app, and chat via QQ direct messages — no public IP required.

Pure Python

More Features

Built for real workflows.

Beyond the basics — workspaces, scheduling, and non-interactive CLI mode for power users.

🗂️

Workspaces

Run multiple fully isolated agents — each with its own config, skills, and memory. Personal agent, work agent, coding agent — all independent, all running simultaneously.

mmclaw run -w ~/.mmclaw_work

⏰

Scheduled Tasks

Just tell your agent what to do and when. Remind me to drink water every 30 minutes. Send me weather at 8am daily. List, modify, or delete schedules anytime in chat.

mmclaw run

⚙️

Non-Interactive CLI

Run a single prompt non-interactively with -p. The agent executes the full agentic loop — tool calls, multi-step tasks — then exits. Clean context every run.

mmclaw run -p "check disk usage"

Quick Start

Zero to agent in 60 seconds.

Two commands. No runtimes, no images, no config files to hunt down. Just pip and run.

pip install mmclaw

mmclaw run

Using Feishu? Install with all extras:
pip install "mmclaw[all]"

terminal

# Step 1 — install
$ pip install mmclaw
✓ Done

# Step 2 — run
$ mmclaw run
Connector: Terminal
LLM: OpenAI
✓ Agent ready