Ralph is an open-source autonomous development loop built for Anthropic's Claude Code. It provides a global CLI that runs continuous Claude Code cycles, with intelligent exit detection, rate limiting, a circuit breaker, response analysis, session continuity, JSON output support, tmux-based live monitoring, and PRD import. Current release is v0.9.3 with 165 passing tests.
Wegent is an open-source AI-native operating system for defining, organizing, and running agentic AI. It provides declarative YAML-driven agent teams, multiple execution engines, isolated sandboxed workspaces, advanced collaboration modes, and integrations for coding workflows and web search.
A demo repository from Anthropic showcasing the Claude Agent SDK. It contains multiple example applications (Email Agent, Excel Demo, Hello World, Research Agent) that illustrate how to build agentic, multi-agent, and AI-assisted workflows with Claude. Intended for local development and learning; not for production. Requires Bun or Node.js 18+ and an Anthropic API key.
DeepAudit is an open-source, multi-agent AI-driven code security auditing platform that combines LLMs, RAG knowledge augmentation, and automated sandbox PoC verification to find and validate vulnerabilities. It supports local Ollama deployments, report generation, and one-click startup via Docker.
A public repository from Anthropic demonstrating Agent Skills for Claude. It contains example skills, the Agent Skills specification, templates, and production-referenced document skills (some Apache 2.0-licensed, some source-available). It shows how to author, install, and use skills in Claude Code, Claude.ai, and via the Claude API.
A2UI is an open-source declarative format and library from Google for agents to generate updateable, framework-agnostic user interfaces as JSON. Clients map A2UI components to native widgets (Web, Flutter, etc.), enabling secure, incremental and portable agent-generated UI rendering. The project is in v0.8 (Public Preview).
Superpowers is a skills library and end-to-end development workflow for coding agents. Built as a plugin/library for agent platforms (notably Claude Code), it provides composable skills that guide an agent through Socratic design refinement, plan generation, test-driven implementation, subagent-driven execution, and structured code review. The system emphasizes TDD, simplicity (YAGNI, DRY), and automated skill-triggering so a coding agent can run with minimal extra instructions.
Deepagents is an open-source agent harness built on LangChain and LangGraph. It equips agents with planning tools, a filesystem backend, sub-agent spawning, and middleware for handling complex, long-horizon tasks reliably and cost-effectively.
Open-source, node-based workflow-automation platform for designing and running complex integrations and AI-powered flows.
Open-source framework that provides composable building blocks to create, orchestrate and monitor LLM-powered applications and agents.
Data framework that connects large-language models to private or enterprise data via indexing, retrieval and agent orchestration.
Memory layer that lets AI agents remember users and context across sessions.
