litellm-agent-platform: self-hosted multi-agent management
litellm-agent-platform manages multiple AI agents efficiently.
Pentest Skill: Python Framework for AI-Orchestrated Web Pentests
Pentest Skill is a Python framework structured as standalone CLI scripts for black-box web penetration testing phases, hosted on GitHub with 11 stars and Python 3.8+ support. It targets AI agents such as Claude Code, Codex, Cursor, or WorkBuddy, enabling full pentest workflows from reconnaissance to
Library-First-Engineering: File-Driven Framework for AI Agent Software
Library-First-Engineering is a process framework that enables developers to build production software using AI agents through file-driven, persona-based workflows. Fully agnostic to languages, IDEs, and LLMs, it prevents common AI pitfalls like logic hallucination by assigning humans high-level thin
slop-review: Native Window for Reviewing AI Agent Code Changes
slop-review is a JavaScript tool hosted on GitHub that provides a dedicated native review window for code changes produced by terminal-based AI coding agents. It supports Claude Code, Codex CLI, and pi by enabling inline comments on diffs via Glimpse and Monaco Editor, with feedback saved to a tempo
oh-my-kimi: Multi-Agent Orchestration Layer for Kimi Code CLI
oh-my-kimi, also known as OMK, offers a production-ready multi-agent orchestration layer atop the Kimi Code CLI (K2.6). It functions as a control plane coordinating Kimi agents for coding tasks through isolated Git worktrees, DAG-based execution plans, evidence-based quality gates, MCP skill hooks,
DeepSeek-Reasonix: Terminal AI Coding Agent with Prefix-Cache Stability
DeepSeek-Reasonix is a terminal-based AI coding agent powered by DeepSeek models that maintains prefix-cache stability across sessions to drastically cut token costs for prolonged coding work. Developers interact via suggested SEARCH/REPLACE operations, applying changes to disk only after manual /ap
Felix: Self-Hosted Gateway for Running AI Agents Locally
Felix delivers a self-hosted gateway for AI agents in a single Go binary that operates entirely on local hardware. It enables connections through CLI or web interface to models like Claude, GPT, Gemini, Qwen, Ollama, or OpenAI-compatible endpoints, supporting in-process tools and remote MCP servers
Self-hosted SprintiQ Brings Agile Planning to AI Coding Agents
SprintiQ is a single-user, self-hosted TypeScript application under Apache 2.0 that provides agile planning for AI-assisted coding workflows. It generates user stories from high-level ideas, manages sprint planning with capacity and velocity tracking, handles task assignments, and maintains bidirect
Medusa: Objective Ranking System for AI Agent Skills
Medusa delivers an automated ranking system that evaluates AI agent skills written as Markdown files named SKILL.md. It measures factors like content length, code blocks, step-by-step instructions, and technical terms to assign one of nine tiers, weighted by 60% complexity, 30% value, and 10% keywor
spring-agent-flow: Framework for Stateful Multi-Agent Workflows in Java
spring-agent-flow is a framework for building stateful multi-agent workflows in Java, layered atop Spring AI. ExecutorAgents handle tasks via chat clients, while CoordinatorAgents route decisions, supporting retries, persistence, shared state, and hybrid AI logic for complex, failure-prone AI system