Most Valuable AI & Open-Source Repositories
A hand-curated directory of battle-tested open-source repositories spanning Claude Skills & MCP servers, local LLM inference engines, autonomous agents, AI code assistants, RAG pipelines, and multimodal tooling.
modelcontextprotocol/servers
Official reference Model Context Protocol (MCP) servers connecting Claude and AI agents to SQLite, Postgres, Filesystems, Git, GitHub, Slack, and Fetch.
modelcontextprotocol/python-sdk
Official Python SDK for the Model Context Protocol, providing FastMCP decorators and asynchronous server/client abstractions.
modelcontextprotocol/typescript-sdk
Official TypeScript and JavaScript SDK for building Model Context Protocol servers and client integrations with type safety.
modelcontextprotocol/inspector
Visual developer testing and debugging client to inspect MCP server tools, resources, and prompt templates in the browser.
modelcontextprotocol/create-python-server
CLI scaffolding utility to bootstrap new production-grade Python MCP projects using FastMCP and modern packaging.
modelcontextprotocol/create-typescript-server
CLI scaffolding tool to quickly generate production-ready TypeScript MCP servers with ESM, types, and build scripts.
anthropics/anthropic-quickstarts
Production-ready reference applications and developer starters for Claude Computer Use, Customer Support, and Financial Analysis.
anthropics/anthropic-sdk-python
Official, production-grade Python client library for the Anthropic Claude API with streaming, tool use, and async support.
anthropics/anthropic-sdk-typescript
Official TypeScript/JavaScript client library for Claude API with full TypeScript types, streaming, and edge runtime support.
anthropics/courses
Interactive hands-on educational notebooks covering Claude prompt engineering, tool use, evaluation, and agentic workflows.
punkpeye/awesome-mcp-servers
Community-curated directory of open-source Model Context Protocol servers, clients, libraries, and developer tools.
wong2/mcp-cli
CLI utility to inspect, test, and invoke Model Context Protocol server tools directly from the terminal.
mark3labs/mcp-go
Full-featured Go implementation of Model Context Protocol for building lightweight, high-speed MCP servers.
executeautomation/mcp-playwright
Playwright browser automation server for Claude and MCP agents to navigate, interact with, and scrape web pages.
simonw/llm-claude-3
Simon Willison LLM CLI plugin adding native Claude 3 and Claude 3.5 support to terminal workflows.
suekou/mcp-notion-server
Model Context Protocol server connecting Claude directly to Notion workspaces, databases, and pages.
automatalabs/mcp-server-docker
MCP server enabling AI agents to safely inspect, start, stop, and inspect logs of Docker containers.
zueai/mcp-manager
Visual desktop management tool to easily install, enable, disable, and configure MCP servers for Claude Desktop.
cloudflare/mcp-server-cloudflare
Model Context Protocol server connecting Claude directly to Cloudflare Workers, KV, D1 databases, and DNS records.
supabase-community/mcp-server-supabase
Model Context Protocol server connecting AI assistants directly to Supabase Postgres databases, Auth, and Storage.
ollama/ollama
Get up and running with Llama 3.3, DeepSeek-R1, Mistral, and Gemma locally with GPU acceleration, CLI, and REST API.
ggerganov/llama.cpp
High-performance LLM inference in pure C/C++ supporting Apple Silicon Metal, NVIDIA CUDA, AVX-512, and GGUF quantization.
vllm-project/vllm
High-throughput, memory-efficient LLM serving engine with PagedAttention, continuous batching, and chunked prefill.
sgl-project/sglang
Fast execution engine and structured decoding framework for complex multi-turn LLM agent workflows and high-speed serving.
turboderp/exllamav2
Ultra-fast inference library for running quantized EXL2 models on modern NVIDIA GPUs with industry-leading tokens/second.
huggingface/text-generation-inference
Production-grade Hugging Face containerized serving toolkit powering Hugging Face Chat and enterprise deployments.
Open-WebUI/open-webui
Feature-rich, self-hosted web UI with RAG, Ollama, OpenAI API compatibility, and multi-user access control.
oobabooga/text-generation-webui
Gradio web UI for running local LLMs via transformers, llama.cpp, ExLlamaV2, and AutoGPTQ with rich extensions.
mudler/LocalAI
Drop-in replacement REST API for OpenAI specifications running on consumer hardware with zero GPU required.
Mozilla-Ocho/llamafile
Distribute and run LLMs with a single portable multi-gigabyte executable file across macOS, Linux, and Windows.
nomic-ai/gpt4all
Open-source desktop chat application and ecosystem to run local LLMs privacy-first on consumer laptops.
janhq/jan
Open-source alternative to ChatGPT that runs 100% offline on your machine with customizable local inference engines.
mlc-ai/mlc-llm
Universal deployment engine that compiles LLMs natively to WebGPU, iOS, Android, and heterogeneous GPUs.
casper-hansen/AutoAWQ
4-bit Activation-aware Weight Quantization (AWQ) engine offering up to 3x inference acceleration with zero perplexity degradation.
deepseek-ai/DeepSeek-V3
State-of-the-art open Mixture-of-Experts (MoE) foundation model architecture and inference code with 671B total parameters.
deepseek-ai/DeepSeek-R1
Open reasoning foundation model trained with large-scale reinforcement learning for complex math, code, and logic tasks.
QwenLM/Qwen2.5
High-performance multilingual foundation models and multimodal vision-language architectures by Alibaba.
huggingface/transformers
State-of-the-art Machine Learning library for PyTorch, TensorFlow, and JAX powering thousands of pretrained foundation models.
unslothai/unsloth
5x faster and 80% less memory LLM fine-tuning library for Llama 3, DeepSeek, Mistral, and Gemma models.
huggingface/peft
Parameter-Efficient Fine-Tuning library enabling efficient adaptation of large pretrained models via LoRA, QLoRA, and Prefix Tuning.
microsoft/DeepSpeed
Deep learning optimization library by Microsoft that makes distributed training and inference scalable and efficient.
Dao-AILab/flash-attention
Fast and memory-efficient exact attention algorithm with IO-awareness for modern NVIDIA GPUs.
browser-use/browser-use
Open-source AI agent that controls web browsers to automate tasks, form filling, research, and interactive navigation.
browserbase/stagehand
High-accuracy AI web browsing framework built on Playwright with natural language actions and schema extraction.
All-Hands-AI/OpenHands
Open platform for autonomous software development agents capable of writing code, running bash, and solving GitHub issues.
Significant-Gravitas/AutoGPT
Pioneering vision and architecture framework for building, testing, and deploying autonomous AI agents.
geekan/MetaGPT
Multi-agent collaborative framework that assigns distinct software engineering roles (architect, engineer, tester) to LLMs.
crewAIInc/crewAI
Production framework for orchestrating role-playing autonomous AI agents with specialized tools and shared memory.
microsoft/autogen
Multi-agent conversation framework enabling complex autonomous and human-in-the-loop task execution.
langchain-ai/langgraph
Library for building stateful, multi-actor applications and cyclical agent graphs with LLMs and human-in-the-loop steps.
elizaOS/eliza
Autonomous multi-agent operating system designed for Discord, Twitter, and multi-platform interactive agents.
stricteq/agent-zero
Transparent, self-contained Python agent that creates its own tools, uses terminal, and writes code autonomously.
assafelovic/gpt-researcher
Autonomous agent that conducts comprehensive online research across hundreds of sources and compiles detailed reports.
mendableai/firecrawl
Web crawler and scraper that turns entire websites into clean markdown ready for LLMs and agents.
unclecode/crawl4ai
High-speed asynchronous open-source web crawler designed specifically for LLM extraction pipelines.
e2b-dev/E2B
Secure cloud sandboxes and execution environments designed specifically for AI agents running untrusted code.
princeton-nlp/SWE-agent
Agent framework that turns LLMs into autonomous software engineers solving real GitHub issues via specialized Agent-Computer Interface.
continuedev/continue
Open-source AI code assistant extension for VS Code and JetBrains supporting local models, custom rules, and docs indexing.
cline/cline
Autonomous coding agent inside VS Code that creates files, runs bash commands, and edits multi-file codebases with human approval.
RooVetGit/Roo-Code
Advanced fork of Cline with custom modes, deep context rules, and MCP tool orchestration for full-stack developers.
paul-gauthier/aider
AI pair programmer in your terminal with Git integration, repo-map generation, and multi-file editing capabilities.
voideditor/void
Open-source AI code editor alternative to Cursor, featuring inline diffs, multi-model support, and local LLM backends.
TabbyML/tabby
Self-hosted AI coding assistant server providing copilot autocompletions and codebase search for engineering teams.
BerriAI/litellm
Unified OpenAI-compatible proxy and SDK to call 100+ LLMs with unified input/output formats, retries, and spend tracking.
boundaryml/baml
Type-safe domain-specific language and framework for structured LLM prompt engineering and output extraction.
instructor-ai/instructor
Structured outputs and schema validation for LLMs powered by Pydantic and Zod with automatic validation retry loops.
dspy-ai/dspy
Framework for algorithmically optimizing LLM prompts and weights instead of manual prompt tweaking.
outlines-dev/outlines
Fast structured text generation guaranteeing valid JSON, regex matches, and context-free grammar compliance at token generation time.
promptfoo/promptfoo
Test and evaluate LLM output quality, security vulnerabilities, prompt regressions, and red-teaming in CI/CD.
astral-sh/uv
Extremely fast Python package installer, resolver, and virtual environment manager written in pure Rust.
astral-sh/ruff
Ultra-fast Python linter and code formatter written in Rust, replacing Black, Flake8, isort, and pyupgrade.
shadcn-ui/ui
Beautifully designed, accessible, copy-pasteable UI components built with Radix UI and Tailwind CSS.
fastapi/fastapi
High-performance Python web framework for building APIs with standard Python type hints and automated OpenAPI docs.
run-llama/llama_index
Comprehensive data framework for building context-augmented LLM applications, RAG pipelines, and structured data agents.
langchain-ai/langchain
Extensive ecosystem and libraries for composing LLM chains, prompts, vector retrievers, and tool integrations.
microsoft/graphrag
Modular graph-based RAG system that builds hierarchical knowledge graphs from unstructured text for holistic dataset reasoning.
infiniflow/ragflow
Open-source RAG engine based on deep document understanding and multi-modal parsing of complex PDFs, Excel, and Word files.
chroma-core/chroma
Open-source AI application database and embedding vector store built for developer experience, speed, and local testing.
qdrant/qdrant
High-performance vector search engine and database written in Rust with extended payload filtering and distributed clustering.
milvus-io/milvus
Cloud-native distributed vector database built for massive-scale similarity search on billions of high-dimensional vectors.
weaviate/weaviate
Open-source AI-native vector database with hybrid search, vector indexing, multi-modal support, and GraphQL API.
pgvector/pgvector
Open-source vector similarity search extension for PostgreSQL supporting exact and approximate nearest neighbor search.
Unstructured-IO/unstructured
Open-source pre-processing library for partitioning, extracting, and cleaning messy unstructured documents (PDFs, PPTX, HTML).
deepset-ai/haystack
Open-source modular framework for building customizable, production-ready search, question-answering, and RAG pipelines.
langfuse/langfuse
Open-source LLM observability, prompt management, tracing, and evaluation platform for production AI applications.
mem0ai/mem0
Intelligent long-term memory layer for AI agents, chatbots, and personalized assistants with graph and vector storage.
FlagOpen/FlagEmbedding
State-of-the-art dense retrieval and reranker embeddings (BGE-M3, BGE-Reranker) for enterprise RAG systems.
dify-ai/dify
Open-source LLM app development platform combining BaaS, orchestration, RAG, and visual workflow builders.
n8n-io/n8n
Fair-code workflow automation platform with native AI agent nodes, vector stores, custom webhooks, and 400+ integrations.
Comfy-Org/ComfyUI
The most powerful modular, node-based GUI and backend for Stable Diffusion, Flux, SDXL, and video diffusion models.
AUTOMATIC1111/stable-diffusion-webui
Widely used web interface for Stable Diffusion text-to-image and image-to-image generation with comprehensive plugins.
ggerganov/whisper.cpp
High-performance C/C++ port of OpenAI Whisper speech recognition model with zero external dependencies.
SYSTRAN/faster-whisper
Optimized reimplementation of Whisper using CTranslate2, achieving up to 4x speedup over PyTorch with lower memory.
suno-ai/bark
Transformer-based text-to-audio model capable of expressive speech, multilingual dialogue, music, and sound effects.
coqui-ai/TTS
Deep learning toolkit for Text-to-Speech synthesis with pre-trained models in 20+ languages and voice cloning.
facebookresearch/segment-anything-2
Meta foundation model for promptable visual segmentation across static images and high-framerate video streams.
livekit/livekit
Ultra-low-latency open-source infrastructure for real-time video, audio, WebRTC, and multimodal voice AI agents.
mifi/lossless-cut
High-speed, lossless video and audio editor and trimming tool built with Electron and FFmpeg with zero quality loss.
PaddlePaddle/PaddleOCR
Multilingual, practical, ultra-lightweight optical character recognition (OCR) toolkit supporting 80+ languages.
haotian-liu/LLaVA
Large Language and Vision Assistant connecting visual encoders with LLMs for open-source multimodal reasoning.
hexgrad/kokoro
Ultra-fast, lightweight 82M-parameter text-to-speech model producing natural English and Japanese speech.
huggingface/diffusers
Pretrained diffusion models for generating images, audio, and 3D structures with PyTorch and modular pipelines.
lllyasviel/ControlNet
Neural network structure to control diffusion models by adding extra conditioning like depth maps, pose estimates, and edge detection.
ultralytics/ultralytics
State-of-the-art computer vision models for object detection, instance segmentation, pose estimation, and tracking (YOLOv8 / YOLO11).
facebookresearch/dinov2
Vision Transformer foundation models producing high-performance visual representations with zero supervision.
openai/clip
Neural network trained on visual concepts and natural language descriptions to enable zero-shot image classification.
cpacker/MemGPT
OS-like memory management system for LLMs with persistent state, context paging, and hierarchical memory tiers (Letta).
simonw/llm
CLI tool and Python library for running prompts against local and cloud LLMs with rich plugin ecosystem.
facebookresearch/faiss
Library for efficient similarity search and clustering of dense vectors on CPUs and GPUs at massive scale.
langflow-ai/langflow
Visual drag-and-drop workspace and rapid prototyping IDE for multi-agent and RAG architectures.
FlowiseAI/Flowise
Drag-and-drop UI to build customized LLM flows, conversational agents, and RAG pipelines with LangChain.js.
activepieces/activepieces
Open-source workflow automation platform with built-in AI integrations, sandboxed code execution, and webhooks.
meilisearch/meilisearch
Lightning-fast, hyper-relevant open-source search engine built in Rust with hybrid vector and keyword search.
typesense/typesense
Fast, typo-tolerant, in-memory open-source search engine optimized for developer ergonomics and vector search.
sashabaranov/go-openai
Unofficial Go client library for OpenAI API including GPT-4, DALL-E, Whisper, and Embeddings.
tiangolo/sqlmodel
Library for interacting with SQL databases from Python code using Python classes and type annotations.
microsoft/typechat
Library that uses TypeScript schemas to build type-safe natural language interfaces with LLMs.
ShishirPatil/gorilla
Fine-tuned LLaMA model that generates accurate API function calls and tool invocations without hallucinations.
vinta/awesome-python
Curated collection of battle-tested Python frameworks, libraries, software, and developer resources.