COMPASS
The Repo Atlas.
202 open-source projects in 19 categories. The tools the workshop builds on, and the ones worth knowing. Every link goes to the source.
Agents · 30
Frameworks and platforms for building autonomous and multi-agent systems.
AutoGen →
Microsoft's multi-agent conversation framework.
AG2 →
Community fork of AutoGen (AgentOS).
CrewAI →
Role-playing collaborative agent crews.
LangGraph →
Stateful, graph-based multi-agent orchestration.
CAMEL →
Communicative agents for scaling role-play.
IoA (Internet of Agents) →
Framework for agents from different ecosystems to collaborate.
SuperAGI →
Dev-first autonomous agent framework.
LangChain →
The foundational LLM application framework.
LlamaIndex →
Data framework for RAG and knowledge agents.
Semantic Kernel →
Microsoft SDK to embed LLMs into apps.
phidata / Agno →
Build agents with memory, tools, and knowledge.
Haystack →
LLM orchestration and RAG pipeline framework.
Odysseus →
Autonomous agent experiment.
AutoGPT →
The original autonomous-agent platform.
BabyAGI →
Minimal task-driven autonomous agent.
MetaGPT →
A multi-agent software company in code.
OpenHands →
Autonomous AI software engineer (ex-OpenDevin).
Agent-Reach →
Agent outreach/automation project.
agency-agents →
Agent agency experiments.
Social-GPT agent →
Autonomous social-media agent.
E2B →
Secure cloud sandboxes for AI agents and code.
ai-agents-frameworks →
Comparison of agent frameworks.
Moltbot →
OpenClaw-family personal assistant bot.
career-ops →
Agent-driven career operations.
karpathy/skills →
Andrej Karpathy's agent skills collection.
Trellis →
Structured agent workflows.
smolagents →
Hugging Face's minimal agent library.
ml-intern →
Hugging Face ML intern agent.
TradingAgents →
Multi-agent LLM financial-trading framework.
pi-mono →
Agent tooling monorepo (Mario Zechner).
Inference & Infrastructure · 7
Serve and run models, from a laptop to a datacenter.
llama.cpp →
The foundational C/C++ local inference engine.
Ollama →
Run open models locally with one command.
vLLM →
High-throughput production inference server.
TensorRT-LLM →
NVIDIA's optimized LLM inference toolkit.
Text Generation Inference →
Hugging Face's production inference server.
Valkey →
Open in-memory key/value store (Redis fork).
DeepSeek-V3 →
Frontier-class open-weight model.
Training & Fine-tuning · 7
Teach models new tricks, fine-tunes, LoRA, and large-scale training.
Unsloth →
Fast, memory-light LoRA fine-tuning.
Axolotl →
Config-driven fine-tuning for many models.
LLaMA-Factory →
Unified fine-tuning for 100+ LLMs.
TRL →
Transformer reinforcement learning (SFT, DPO, PPO).
DeepSpeed →
Microsoft's large-scale training optimizer.
PEFT →
Parameter-efficient fine-tuning methods.
ColossalAI →
Toolkit for large-model parallel training.
Evaluation · 6
Measure whether the model actually works.
lm-evaluation-harness →
The standard open LLM benchmark harness.
promptfoo →
Test and evaluate prompts and models.
HELM →
Stanford's holistic model evaluation.
Chatbot Arena →
LMSYS human-preference model ranking.
BIG-bench →
Google's collaborative benchmark suite.
chatbot_arena_conversations →
Open dataset of Arena battles.
MCP (Model Context Protocol) · 9
Servers and tools for the open protocol that connects agents to everything.
MCP servers →
The official reference MCP server collection.
n8n-mcp →
MCP server exposing n8n workflows.
higgsfield-mcp →
MCP server for Higgsfield.
blender-mcp →
Drive Blender from an AI agent via MCP.
claude-mem →
Persistent memory for Claude via MCP.
anthropic-mcp-servers →
A collection of Anthropic MCP servers.
claude-context →
Codebase context for Claude via MCP.
awesome-mcp-servers →
Curated list of MCP servers.
roundtable →
Multi-agent MCP roundtable.
Vector & RAG · 7
Store embeddings and retrieve knowledge for grounded answers.
Chroma →
Developer-friendly open embedding database.
Qdrant →
High-performance Rust vector database.
Weaviate →
Open vector database with hybrid search.
Milvus →
Cloud-native vector database at scale.
pgvector →
Vector search inside PostgreSQL.
PageIndex →
Reasoning-based document indexing for RAG.
RAGFlow →
Deep-document-understanding RAG engine.
Observability · 6
See what your agents and models are actually doing.
Langfuse →
Open-source LLM tracing and analytics.
LangSmith →
LangChain's tracing and eval platform.
Helicone →
Open observability proxy for LLM calls.
Weights & Biases →
Experiment tracking for ML/AI.
Netdata →
Real-time infrastructure monitoring.
agent-orchestrator →
Composio's agent orchestration layer.
Automation · 5
Wire tools together, visual workflow builders and low-code glue.
n8n →
Fair-code workflow automation with AI nodes.
Flowise →
Drag-and-drop LLM app builder.
Langflow →
Visual builder for LangChain flows.
Activepieces →
Open-source no-code automation.
Node-RED →
Flow-based programming for events and IoT.
Coding Agents & Tools · 16
AI that writes, reviews, and runs code.
GitHub Copilot →
GitHub's AI pair programmer.
Cursor →
The AI-first code editor.
Cody →
Sourcegraph's codebase-aware AI assistant.
Continue →
Open-source AI code assistant for any IDE.
Cline →
Autonomous coding agent in your editor.
Codex →
OpenAI's coding agent CLI.
Aider →
AI pair programming in your terminal.
Open Interpreter →
Natural-language code execution locally.
CopilotKit →
Build in-app AI copilots and agents.
codegraph →
Graph-based code understanding.
Claude Code →
Anthropic's official coding agent CLI.
kimi-cli →
Moonshot Kimi coding CLI.
pixel-agents →
Agent tooling for builders.
bloop →
Fast code search with natural language.
Gemini CLI →
Google's Gemini coding CLI.
Goose →
Block's open, extensible AI agent.
Creative & Media · 21
Image, video, audio, and voice generation.
ComfyUI →
Node-based diffusion workflow engine.
Stable Diffusion WebUI →
The classic image-generation UI.
FLUX →
Black Forest Labs' image models.
Stable Diffusion →
Stability AI's diffusion models.
HunyuanVideo →
Tencent's open video generation model.
HunyuanVideo 1.5 →
Next-gen Hunyuan video model.
HunyuanVideo-I2V →
Image-to-video from Hunyuan.
NVIDIA Cosmos →
World-foundation models for physical AI.
VideoCrafter2 →
Open text-to-video generation.
LTX-Video →
Lightricks' fast video generation.
Remotion →
Make videos programmatically in React.
Awesome-Suno →
Curated Suno AI music resources.
suno-ai →
Suno AI music tooling.
Bark →
Suno's text-to-audio / TTS model.
Amphion →
Open audio, music, and speech toolkit.
MoneyPrinterTurbo →
Auto-generate short videos with AI.
Qwen3-TTS →
Alibaba Qwen text-to-speech.
Qwen3-ASR →
Alibaba Qwen speech recognition.
VALL-E Voice →
Microsoft neural codec voice model.
stable-worldmodel →
Open world-model research.
Pixelle-Video →
AIDC-AI video generation.
Multimodal · 4
Models that see as well as read.
LLaVA →
Large language-and-vision assistant.
Qwen-VL →
Alibaba's vision-language models.
InternVL →
Open multimodal foundation models.
comfyui_qwen2vl-flux →
Qwen2-VL + FLUX ComfyUI nodes.
Private / Local Assistants · 6
Own-your-data chat and assistants that run on your hardware.
GPT4All →
Run local chatbots on consumer hardware.
LocalAI →
Drop-in local OpenAI-compatible API.
Jan →
Offline, open-source ChatGPT alternative.
llamafile →
Run an LLM as a single executable file.
privateGPT →
Ask questions of your documents offline.
h2oGPT →
Private local GPT with document Q&A.
AI Security & Guardrails · 6
Keep models and agents safe, aligned, and un-jailbroken.
NeMo Guardrails →
NVIDIA toolkit for LLM guardrails.
LLM Guard →
Security toolkit for LLM interactions.
aiguard-nemo-guardrails →
Zscaler AI guardrails integration.
airs-nemo-guardrails →
Palo Alto AI runtime security.
Semgrep →
Fast static analysis for finding bugs.
bumblebee →
Perplexity security tooling.
Quantization & Optimization · 5
Make models smaller and faster.
bitsandbytes →
8-bit and 4-bit quantization for PyTorch.
AutoGPTQ →
GPTQ quantization toolkit.
AutoAWQ →
Activation-aware weight quantization.
HQQ →
Half-quadratic fast quantization.
BitNet →
Microsoft's 1-bit LLM inference.
Prompting & Programming LLMs · 4
Move beyond prompt strings to structured, testable programs.
DSPy →
Program your language models rather than prompting them.
Guidance →
Constrained generation and control.
prompttools →
Test and experiment with prompts.
Prompt Engineering Guide →
The reference guide to prompting.
Learning & Awesome Lists · 25
Where to start and how to go deep, curated lists, books, and courses in code.
awesome →
The awesome list of awesome lists.
Awesome-LLM →
Curated LLM papers, models, and tools.
awesome-agents →
Curated AI-agent projects.
Awesome-AI-Agents →
Another strong agents list.
awesome-ai-agents →
Agent frameworks and resources.
Awesome-LLMOps →
Curated LLMOps tooling.
awesome-llm-and-aigc →
LLM and generative-AI resources.
awesome-python →
Curated Python frameworks and libraries.
awesome-machine-learning →
Curated ML frameworks by language.
awesome-datascience →
Data-science learning roadmap.
awesome-mlops →
Curated MLOps references.
awesome-generative-ai-guide →
GenAI courses, papers, and interviews.
developer-roadmap →
Roadmaps to becoming a developer.
PythonDataScienceHandbook →
Jake VanderPlas's full book in notebooks.
handson-ml3 →
Géron's Hands-On ML, 3rd edition.
dmls-book →
Chip Huyen's Designing ML Systems.
nn-zero-to-hero →
Karpathy: build neural nets from scratch.
minGPT →
A minimal, readable GPT implementation.
nanochat →
Karpathy's minimal ChatGPT clone.
project-based-learning →
Learn by building real projects.
build-your-own-x →
Recreate the tech you love, from scratch.
Made-With-ML →
Design, build, and deploy ML in production.
awesome-indie-hackers-tools →
Tools for solo builders.
indie-dev-toolkit →
Curated indie-developer toolkit.
awesome-ai-agents-2026 →
2026 agents roundup.
Scraping & Data Collection · 3
Feed your agents the web.
Scrapling →
Adaptive, stealthy Python web scraping.
Crawlee (Python) →
Reliable crawling and scraping library.
Firecrawl →
Turn websites into LLM-ready data.
Utilities & Tools · 28
The odds and ends that make everything else easier.
MarkItDown →
Convert files to Markdown for LLMs.
notebooklm-py →
NotebookLM-style workflows in Python.
browser-use →
Let agents control a real browser.
taste-skill →
An agent 'taste' skill.
openai/plugins →
OpenAI plugin examples.
pm-skills →
Product-management agent skills.
last30days-skill →
Recent-activity agent skill.
ui-ux-pro-max-skill →
UI/UX design agent skill.
superpowers →
A framework of agent skills.
open-notebook →
Open-source NotebookLM alternative.
obsidian-releases →
Obsidian plugins and themes registry.
apple/container →
Apple's Linux container runtime for macOS.
rilable →
Builder utility.
tolaria →
Refactoring tooling.
ian-xiaohei-illustrations →
Open illustration set.
headroom →
Context/'headroom' utility.
impeccable →
Frontend design-quality skill.
heretic →
Model-abliteration research tool.
OpenViking →
Volcengine open project.
MiroFish →
Open utility project.
caveman →
Minimalist dev tool.
ECC →
Engineering command-center agent toolkit.
Factory →
Agent-driven software delivery.
liteparse →
Lightweight document parsing (LlamaIndex).
beautiful-mermaid →
Nicer Mermaid diagram rendering.
Twenty →
Open-source CRM.
DeepSeek-OCR-2 →
Open OCR model.
memU →
Memory layer for agents.
Platforms & UIs · 7
Full products and hubs you can self-host or build on.
Hugging Face Hub →
The hub client for models and datasets.
Replicate →
Run open models with one API call.
Supabase →
Open-source Firebase alternative.
v0 →
Vercel's generative UI tool.
Lobe Chat →
Polished open-source AI chat UI.
Open WebUI →
Self-hosted UI for local models.
AnythingLLM →
All-in-one private RAG desktop app.