freshcrate
Skin:/
Home > Frameworks > ag-claw

ag-claw

Modular AI agent framework with 59 pluggable features, 8+ messaging channels, and production-grade security. TypeScript-first. MIT license. Self-hosted, no subscriptions.

Why this rank:Recent releaseHealthy release cadenceStrong adoption

Description

Modular AI agent framework with 59 pluggable features, 8+ messaging channels, and production-grade security. TypeScript-first. MIT license. Self-hosted, no subscriptions.

README

AG-Claw

Modular AI agent framework. Self-hosted. No subscriptions.

VersionNodeLicenseCIStarsTelegramDockerDockerHub


# Clone and run
git clone https://github.com/AG064/ag-claw.git
cd ag-claw
npm install
npm run onboard
npm run dev

npm run onboard launches the interactive setup wizard and writes your first agclaw.json.

npm run dev starts AG-Claw using the saved configuration. If you skip the wizard, AG-Claw will boot with defaults and immediately start the server/webchat.

Or use Docker:

docker run -it ag064/ag-claw

Or grab a binary from the latest release.


What Is This?

AG-Claw is an AI agent framework built on top of OpenClaw. It gives you a modular system where you enable exactly the capabilities you need — Telegram bots, memory backends, automation tools, security layers, and more.

Your data stays on your machine. No cloud dependency. No subscriptions.


What's Inside

Communication: Telegram, Discord, Slack, WhatsApp, Email (IMAP/SMTP), SMS, Webchat, Signal.

Memory: SQLite, semantic search, knowledge graph, hierarchical memory, git sync.

Automation: Cron scheduler, mesh workflows, file watcher, webhooks, browser automation, container sandbox.

Security: Encrypted secrets, rate limiting, allowlists, policy engine, audit logging.

Tools: AI image generation with fallback chains, YouTube shorts processing, skill loader, self-improving agent.


Why AG-Claw

You get 65+ features as plugins. Toggle what you need, ignore the rest. Everything runs locally. No vendor lock-in.

What you get Other frameworks
65+ ready features You build everything from scratch
8 communication channels Usually one, if any
5 memory backends Usually one
TypeScript throughout Often JavaScript or wrappers
Security-first design Security as an afterthought
Docker ready Often requires manual setup

Installation Options

From source (recommended for development):

git clone https://github.com/AG064/ag-claw.git
cd ag-claw
npm install
npm run onboard   # optional, but recommended for first-time setup
npm run build
npm start

If you want the interactive onboarding wizard without building first, use:

npm run onboard

npm start starts the server/runtime, not the setup wizard.

Docker (recommended for production):

docker run -it ag064/ag-claw

Binary (no dependencies): Download from github.com/AG064/ag-claw/releases/latest — pick the one for your OS:

chmod +x agclaw-*
./agclaw-* --help

npm (coming soon):

npm install -g ag-claw

CLI Commands

agclaw init                    # Initialize in current directory
agclaw onboard                 # Interactive setup wizard
agclaw gateway start           # Start gateway
agclaw gateway stop            # Stop gateway
agclaw gateway status          # Check if running
agclaw gateway logs           # View logs

agclaw tools                   # List all features
agclaw feature <name> enable   # Enable a feature
agclaw feature <name> disable # Disable a feature

agclaw agents                  # List configured agents
agclaw sessions                # View conversation sessions
agclaw memory search <query>   # Search memory
agclaw memory stats            # Memory statistics

agclaw config                  # Show config
agclaw config <key>           # Get specific value
agclaw config <key> <value>   # Set a value

agclaw doctor                  # Diagnose issues
agclaw connect                 # Setup integrations

Architecture

AG-Claw Gateway
├── Channels (Telegram, Discord, Webchat...)
├── Features (65+ plugins)
│   ├── computer-control
│   ├── image-generation
│   ├── skill-evolution
│   ├── knowledge-graph
│   └── ...59 more
├── Agentic Tool Loop
│   ├── LLM Provider (OpenRouter, Anthropic, Ollama...)
│   ├── Tools (registered by features)
│   └── Memory (semantic, graph, SQLite)
└── Security Layer
    ├── Audit Log
    ├── Rate Limiting
    ├── Allowlists / Denylists
    └── Policy Engine

Remote Access

SSH tunnel (recommended for local networks):

ssh -L 3000:localhost:3000 user@your-server
# open http://localhost:3000

Tailscale (VPN, works anywhere):

tailscale up
tailscale serve https

Cloudflare Tunnel (zero config):

cloudflared tunnel --url http://localhost:3000

Documentation

Guide What it's for
Quick Start Up and running in 5 minutes
User Guide Operating AG-Claw day to day
Developer Guide Adding features and contributing
API Reference REST endpoints and config schema
Migration Guide Switching from OpenClaw
Security Security features and best practices
Features All 65 features documented

Contributing

git clone https://github.com/AG064/ag-claw.git
cd ag-claw
npm install
npm run build
npm test
git checkout -b feature/your-feature-name

git clone https://github.com/AG064/ag-claw.git cd ag-claw npm install npm run build npm test git checkout -b feature/your-feature-name


---

## License

MIT. Copyright 2024-2026 AG064. Based on OpenClaw by nickarora.

---
MIT. Copyright 2024-2026 AG064. Based on OpenClaw by nickarora.

---

<p align="center">
Questions? Open an issue on <a href="https://github.com/AG064/ag-claw/issues">GitHub</a>.
</p>
<p align="center">
Questions? Open an issue on <a href="https://github.com/AG064/ag-claw/issues">GitHub</a>.
</p>

Release History

VersionChangesUrgencyDate
v0.0.7# Argentum v0.0.7 v0.0.7 is a desktop polish release. It keeps the security cleanup already merged into `development`, then focuses on the parts of the app people touch first: chat, onboarding, provider setup, Telegram session handling, and app-visible diagnostics. This release also adds the first stable Argentum llama.cpp local server path. The Windows setup executable can install the local server binaries when selected in the installer, while the default install remains lightweight. Users whHigh5/23/2026
v0.0.5# Argentum v0.0.5 Argentum v0.0.5 is the first release pass focused on making the desktop MVP feel like a coherent product: stable ChatGPT/OpenAI and MiniMax provider paths, MiniMax usage visibility, cleaner gateway output, reasoning-block display controls, tighter chat history behavior, and explicit OS/hardware support notes. ## Install Recommended downloads: - `Argentum_0.0.5_x64-setup.exe` - Windows desktop setup executable. - `Argentum_0.0.5_x64_en-US.msi` - direct Windows Installer packHigh5/8/2026
v0.0.4# Argentum v0.0.4 Argentum v0.0.4 fixes the Windows release packaging problem where the prominently named installer assets launched the terminal CLI instead of the desktop interface. This release makes the GUI installer the default user path and keeps CLI binaries explicitly optional. ## Install Recommended downloads: - `Argentum_0.0.4_x64-setup.exe` - Windows desktop setup executable. - `Argentum_0.0.4_x64_en-US.msi` - direct Windows Installer package. - `Argentum_0.0.4_amd64.AppImage`, `.dHigh4/29/2026
v0.0.2Latest release: v0.0.2High4/26/2026
v0.0.1Latest release: v0.0.1High4/14/2026
v0.4.0## Get Started in 30 Seconds **From source:** ```bash git clone https://github.com/AG064/ag-claw.git cd ag-claw npm install npm run build npm start ``` **Docker (recommended):** ```bash docker run -it ag064/ag-claw ``` **Binary (no dependencies):** Download from the Assets section below. Pick the file for your OS, then: ```bash chmod +x agclaw-* ./agclaw-* --help ``` --- ## What's New in v0.4.0 This is a major production release with 147 commits of new features, security hardening, and CI/High4/6/2026

Dependencies & License Audit

Loading dependencies...

Similar Packages

DreamServerLocal AI anywhere, for everyone — LLM inference, chat UI, voice, agents, workflows, RAG, and image generation. No cloud, no subscriptions.v2.5.3
agent2The production runtime for AI agents. Schema in, API out. Built on PydanticAI + FastAPI.v0.4.0
astack🤖 A composable framework for building AI applications.v0.1.1-beta.0
mayrosProduction-ready AI agent framework — semantic memory, multi-agent mesh, MCP server, intelligent routing, governance, and 67+ platform integrations.v0.3.2
outputThe open-source TypeScript framework for building AI workflows and agents. Designed for Claude Code describe what you want, Claude builds it, with all the best practices already in place.main@2026-06-05

More in Frameworks

spec_driven_developSpec-Driven Develop is a platform-agnostic AI agent skill that automates the pre-development workflow for large-scale complex tasks. It is not a framework, not a runtime, not a package manager — it is
deer-flowAn open-source long-horizon SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of ta
tqdmFast, Extensible Progress Meter
simBuild, deploy, and orchestrate AI agents. Sim is the central intelligence layer for your AI workforce.