🦞 OpenClaw

Your personal AI assistant. Any OS. Any platform. The lobster way.

OpenClaw is a fully personal AI assistant that runs on your own devices. It integrates directly into the communication channels you already use — including WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, Microsoft Teams, and more.

Designed as a single-user, always-on assistant, OpenClaw feels local, fast, and persistent. It supports voice interaction across macOS, iOS, and Android, and even provides a live Canvas interface you control. The Gateway acts as the control plane, while the assistant itself is the core product.

Category: Personal AI Assistant / Multi-Channel Agent

🧠 Pi Monorepo (by badlogic)

AI agent toolkit for developers

Pi Monorepo is an open-source AI agent toolkit designed to help developers build, run, and deploy intelligent agents powered by large language models (LLMs).

Rather than being a single tool, it is a structured monorepo containing multiple interconnected packages. These include a unified multi-provider LLM API, an agent runtime, a coding agent CLI, terminal and web UI libraries, and deployment tools for managing vLLM on GPU infrastructure.

Together, they provide a full-stack foundation for building production-ready AI agents.

Category: Agent Framework / Infrastructure Toolkit

🔐 Shannon (by KeygraphHQ)

Autonomous AI hacker for real exploit discovery

Shannon is an AI-powered penetration testing agent that goes beyond static analysis and alerts. Instead of simply flagging potential vulnerabilities, it autonomously executes real exploits to prove that vulnerabilities are actually exploitable.

It analyzes source code, identifies potential attack vectors, and uses its built-in browser to simulate real-world user behavior. Shannon performs attacks such as injection exploits and authentication bypass to validate risk.

With a reported 96.15% success rate on the hint-free, source-aware XBOW benchmark, Shannon represents a new class of autonomous security testing agents.

Category: AI Security / Autonomous Pentesting

📚 QMD (by tobi)

Local-first hybrid search engine for your knowledge

QMD (Query Markup Documents) is a fully local hybrid search engine built for personal knowledge bases. It indexes Markdown notes, meeting transcripts, and documentation, enabling both keyword-based and semantic search.

QMD combines BM25 full-text search, vector embeddings, and LLM-based re-ranking — all running locally through GGUF models via node-llama-cpp. Designed specifically for agentic workflows, it provides high-quality retrieval without relying on external APIs.

This makes it ideal for powering local RAG systems and AI agents that need reliable, private knowledge retrieval.

Category: Local RAG Engine / Agent Retrieval Layer

🧩 Agent Skills (openai/skills)

Reusable capabilities for AI agents

Agent Skills is a standardized capability system for AI agents. It allows instructions, scripts, and resources to be packaged into reusable, task-specific modules.

Built for Codex, it enables teams to install, distribute, and reuse skills across environments. Instead of repeatedly writing prompts or reconfiguring tools, developers can create a skill once and reuse it everywhere.

It effectively introduces a plugin ecosystem model for AI agents.

Category: Agent Plugin System / Capability Standard