Open source software directory ranked & explained
Discover, compare, and evaluate open-source projects in under a minute — live trending, health scores, alternatives, and AI summaries. No static lists.
Rising this week
216
Total stars
16.5M
skills
Shell
Skills for Real Engineers. Straight from my .claude directory.
Weekly momentum
View all- 1
skills
Skills for Real Engineers. Straight from my .claude directory.
+15.5k
197.6k stars
- 2
agency-agents
A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.
+5.3k
135.8k stars
- 3
superpowers
An agentic skills framework & software development methodology that works.
+5.0k
264.4k stars
- 4
ponytail
Ponytail injects a rule‑set into AI agent sessions that checks each proposed change against a ladder: need, reuse, stdlib, native feature, dependency, one‑line, then minimum viable code. Benchmarks show ~54% less code, ~20% lower cost, ~27% faster execution, and 100% safety compared to a no‑skill baseline. Installation is a single marketplace add for Claude Code, Codex, Copilot CLI, OpenCode, Gemini/Antigravity, Hermes, CodeWhale, Swival, Devin, and OpenClaw.
+4.5k
80.9k stars
- 5
hermes-agent
The agent that grows with you
+4.4k
223.3k stars
- 6
firecrawl
The API to search, scrape, and interact with the web at scale. 🔥
+4.2k
158.6k stars
Editor's picks
View alln8n
TypeScriptSelf-hosted workflow automation with native AI capabilities and 400+ integrations.
ollama
GoRun open LLMs locally with one command. CLI and REST API included.
langchain
PythonThe most widely used framework for building LLM applications with chains, agents, and RAG.
Flowise
TypeScriptVisual low-code builder for LangChain flows, RAG chatbots, and AI agents.
Community favorites
View allbuild-your-own-x
MarkdownMaster programming by recreating your favorite technologies from scratch.
awesome
😎 Awesome lists about all kinds of interesting topics
public-apis
PythonA collective list of free APIs
freeCodeCamp
TypeScriptfreeCodeCamp.org's open-source codebase and curriculum. Learn math, programming, and computer science for free.
free-programming-books
Python:books: Freely available programming books
openclaw
TypeScriptYour own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞
Explore by category
View allDeveloper Tools
Languages, SDKs, and dev productivity
234 projects
AI Agents
Autonomous agents, orchestration, and agent frameworks
137 projects
AI
AI agents, LLM frameworks, and machine learning tools
43 projects
Self-Hosted
Run on your own infrastructure
39 projects
Claude Code
Open-source projects tagged with Claude Code
34 projects
DevOps
CI/CD, containers, and infrastructure tools
22 projects
Codex
Open-source projects tagged with Codex
21 projects
API Frameworks
Backend frameworks and API tooling
20 projects
Curated collections
View allLangChain vs n8n vs RAGFlow: LLM Framework, Workflow Automation, and RAG Engine Comparison
Compare LangChain, n8n, and RAGFlow on features, deployment, learning curve, and best‑fit use cases to choose the right AI tool for your project.
3 projects
n8n vs Appsmith: Workflow Automation vs Internal Tool Builder Comparison
Compare n8n’s AI‑native workflow automation with Appsmith’s low‑code internal tool builder—features, licensing, deployment, and best‑fit use cases.
2 projects
Recently added
View allinterior
TypeScriptInterior provides high-fidelity, physics-based micro-interactions that prioritize accessibility, keyboard support, and smooth state transitions. It follows a 'copy-and-paste' philosophy where components are provided as single files containing a headless hook for logic and a styled component for visual reference, ensuring developers can easily reskin components without losing complex behavioral logic.
ai-sdk
GoThe Grafana AI SDK for Go enables developers to build AI-powered applications by providing a standardized interface for multiple providers (Anthropic, OpenAI, Bedrock, etc.). It allows Go backends to communicate seamlessly with React frontends via Server-Sent Events (SSE), supporting features like structured output, tool-calling, and agent observability.
hwatu
RustHwatu delivers a 35 ms verification pass (load, eval, screenshot) with one command, 97 % pixel‑match accuracy, headless or headed windows, ad‑blocking, animation metrics, console and request error reporting, and interruptible human hand‑off—all from a single static binary that relies on the host’s WebKitGTK. It exposes its functionality via a minimal MCP server, a CLI, or a one‑line JSON socket, making it easy to integrate into existing agent harnesses.
sakura-crossing
JavaScriptSakura Crossing is a procedural 3D world built with Three.js and Vite that simulates a Japanese suburb in blossom season. It features a unique '3D-to-2D' aesthetic achieved through custom toon materials, screen-space ink lines, and inverted-hull outlines. The world is mapped onto a spherical planet (160m radius) and includes complex features like a seamless railway, a drainage channel, and procedurally generated hills and water bodies, all without a single image asset in the source directory.
verified-3d-mesh-intersection
LeanThe project offers a Lean 4 verified kernel for exact 3D mesh intersection that satisfies a formal specification linking the output solid to the intersection of input solids and ensures well-formedness properties. The kernel is produced by AI agents, but its correctness is guaranteed by 60k lines of machine-checked proofs; a reviewer only needs to examine the 93-line specification and run Lean's checker. A WebAssembly demo enables interactive experimentation, though the UI and glue code remain unverified.
optim-plans
Pythonoptim-plans offers a command‑line controller and skill interface for Claude and Codex that: - Inspects a target repository and surfaces improvement opportunities. - Generates a PLAN_v1.md, then iteratively refines it with reviewer or criticizer loops. - Produces a manifest‑bound execution plan, runs code in a single controller‑owned worktree, and validates output with a read‑only validator. - Stores public artifacts as Markdown in docs/optim-plans/YYYY‑MM‑DD‑topic/ and machine state under .git/optim-plans/ for replayability. - Enforces safety via immutable manifests, nonce‑based approvals, Git audits, and bounded retry logic. - Supports both local CLI and plugin usage for Claude and Codex, with configurable worker choices and effort levels.