> 比 Claude Cowork 更底层、更自主的 AI 智能体系统

> Read this in other languages: Русский

AI Enhanced Binary Editor (v0.1)

zjwangalauda__Alauda-GROS-Copilot

Risk analytics for Kalshi prediction market portfolios. Computes VaR/CVaR, liquidity metrics, factor decomposition, optimal position sizing, and scenario stress tests.

zhangchuman1029-debug__digital-bonfire

基于真实行业数据的陪诊业务沙箱模拟环境，用于 GMV 预测、运营决策支持和业务参数验证。

Browser-based control for Thorlabs instruments via REST API and real-time streaming.

A document processing daemon that watches a directory for new PDFs and images, extracts content via parallel OCR + Vision LLM, classifies documents by recipient and category, and indexes everything into a RAG system for chat-based retrieval.

zaidy7800__langchain-chatbot

An MCP server that connects AI assistants to your Zotero library — including full-text PDF extraction. Search papers, browse collections, pull citations, and read the actual content of your PDFs, all from within Claude Code or any MCP-compatible client.

A persistent, expressive AI companion for your desktop. Talk to an AI character with personality, memory, and voice — running locally with Ollama or via cloud providers.

AI-Powered Integration Engine for B2B Software Solutions

> FastAPI service wrapping Google NotebookLM via notebooklm-py for notebook management, querying with citations, and Zotero source integration.

> Lightweight service serving analytical definitions without execution logic

PDF-based knowledge management system with RAG (Retrieval Augmented Generation). Upload PDFs or Markdown documents, build vector embeddings in Qdrant, and generate custom content (quotes, posts, summaries, insights) using Claude AI with anti-hallucination controls.

Multi-source web scraping and enrichment pipeline for building comprehensive founder profiles. Aggregates data from articles, blogs, podcasts, YouTube videos, LinkedIn, and press mentions into structured Markdown profiles with optional LLM-powered synthesis.

Hugo 블로그 관리를 위한 API 서버와 MCP 클라이언트입니다.

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

> 📚 推荐阅读：如果你关注的是"如何从真实问题到系统设计"，建议先阅读 方法论文档。

xnotwostar__ai-news-radar

A fully automated research system for quantitative finance that orchestrates the entire research lifecycle—from literature review to experiment execution to report generation—using a multi-agent collaboration architecture powered by Claude AI.

Standardized Model Development and Deployment Platform

Live bilingual (English/Spanish) speech-to-text for church outreach at Stark Road Gospel Hall, Farmington Hills, MI.

A production-ready Retrieval-Augmented Generation (RAG) system that answers questions based on your document knowledge base. Deployed on AWS with serverless architecture.

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

A template to develop a group of Python Applications by using uv and polylith.

Five thousand years of writing. Less than 2% translated.

Mission-driven hypothesis testing organism: agents collaborate via shared message board, vote on evolution, earn credits through KPI success.

Domestic AI Cloud Manufacturing Platform.

A self-hosted AI platform with LLM chat, image generation, and a Claude Code-like coding assistant. No cloud, no subscriptions - runs entirely on your hardware.

A Model Context Protocol (MCP) server that gives Claude (or any MCP client) direct access to local Microsoft OneNote notebooks via the COM API. No Azure registration, no API keys, no cloud auth — just your local OneNote desktop app.

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

A video download, transcription, translation, and encoding microservice.

What we're building: A bidirectional sync engine between Airtable and Priority ERP covering 9 workflows across products, vendors, and customers. Tech stack: Pure Python (no n8n). Hosted on Railway. User context: The project owner is not a developer. Keep explanations clear, ask befor

LLM-powered entity extraction + hybrid RAG (vector + graph) from Griffith's Ramayana translation.

A Nix Flake for running ComfyUI on Linux with direnv support.

vivid-__BTC-signal

A Figma plugin that sends design screenshots to AI personas for instant, multi-perspective design feedback. Select frames in Figma, pick reviewer personas (e.g., "First-Time User", "Accessibility Advocate"), and get structured feedback with annotated issue overlays streamed back in real time.

AI Emergency Response System — Automated ePCR for Paramedics.

Vijil Dome is a fast, lightweight, and highly configurable library for adding runtime guardrails to your AI agents. It combines top open-source LLM safety tools with proprietary Vijil models to detect and respond to unsafe content — with built-in support for observability, tracing, and popular a

vhsrewinder__onq-lidar

veraymw__quinn-learning-service

A ~3,700 LOC minimal reimplementation of TensorRT-LLM's PyTorch backend, built to teach the architectural patterns that distinguish TRT-LLM from vllm and sglang.

> Research prototype. LEKS is under active development. The core architecture is implemented and functional, but it has not been validated at scale. APIs may change. Do not use in production without thorough evaluation against your specific workload. See Evaluation for the benchma

A reference implementation for LLM + RAG development, using Telegram as a convenient UI layer.

Interactive force-directed graph visualization of Brandon Sanderson's Cosmere, built from 16,000+ Words of Brandon Q&A entries from Arcanum.

A comprehensive FastMCP server providing access to the Multilead Open API with 74 endpoints for lead management, campaigns, conversations, webhooks, and analytics.

Fluent builder API for Google's Agent Development Kit (ADK). Reduces agent creation from 22+ lines to 1-3 lines while producing identical native ADK objects.

This Quarto book uses LLMs to evaluate research papers based on The Unjournal metrics, and compares the results to human evaluations.