Denoise

This category shows a clear shift from simple prompt collections to structured, reusable "skills" for agents. Contributions from well-known developers like Matt Pocock and Addy Osmani signal a move towards professionalizing agent development.

• mattpocock/skills

  From a well-known TypeScript educator, this repository defines a structured format for 'Agent Skills'. Its popularity signals a move towards creating reusable, well-documented capabilities for AI agents, moving beyond simple ad-hoc prompts.

  rising⭐ 28k
• forrestchang/andrej-karpathy-skills

  This is a single-file prompt based on Andrej Karpathy's analysis of LLM coding patterns. Its high ranking for a simple markdown file demonstrates the immense value the community places on expert-curated instructions for controlling AI agent behavior.

  repeated⭐ 93k
• ComposioHQ/awesome-codex-skills

  A curated list of skills for the Codex platform. As an 'awesome list', its trendiness is more a reflection of the growing ecosystem around Codex than a new technology itself. It serves as a discovery tool for platform users.

  new⭐ 2.4k
• luongnv89/claude-howto

  A practical, visual guide to using Claude Code. Its sustained popularity indicates a strong and ongoing need for accessible, hands-on documentation to help developers get productive with complex new AI tools.

  cooling⭐ 29k
• addyosmani/agent-skills

  From a prominent engineer at Google, this project also focuses on production-grade skills for AI agents. Its emergence alongside `mattpocock/skills` solidifies the 'skills' abstraction as a key trend in maturing agent development.

  new⭐ 24k

The focus here is on access and extension. Community projects like `Alishahryar1/free-claude-code` provide wrappers to proprietary models, while others like `Yeachan-Heo/oh-my-codex` add extensibility, indicating gaps in the official offerings.

• Alishahryar1/free-claude-code

  This tool provides free command-line and VSCode access to the otherwise proprietary Claude Code agent. It's trending because it directly addresses the access and cost barriers for developers wanting to experiment with a state-of-the-art AI coding tool.

  rising⭐ 15k
• Yeachan-Heo/oh-my-codex

  An extension framework for the Codex agent, allowing users to add hooks, create agent teams, and more. This reflects a desire to customize and build on top of base agent platforms, similar to how shell customization frameworks evolved.

  cooling⭐ 26k
• davila7/claude-code-templates

  A command-line tool for managing Claude Code configurations. This is a typical utility that emerges in a maturing ecosystem, designed to reduce boilerplate and simplify the developer workflow for a specific platform.

  new⭐ 26k
• CJackHwang/ds2api

  This middleware converts the DeepSeek client protocol into a universal API format. It's a technical solution aimed at interoperability and abstracting away vendor-specific implementations, a common need in a fragmented market.

  new⭐ 1.7k
• rtk-ai/rtk

  A command-line proxy written in Rust that aims to reduce LLM token usage. This tool tackles a significant operational pain point for developers: the cost of frequent API calls. Its focus on efficiency is highly practical.

  cooling⭐ 37k

The single trending repository, `Z4nzu/hackingtool`, is a common pattern for this category: an aggregated collection of existing scripts and tools. Its popularity reflects persistent interest in security tooling, but it is not a novel technical development.

• Z4nzu/hackingtool

  An all-in-one collection of scripts for hacking and security tasks. Such repositories are perennially popular but are typically aggregations of existing tools rather than novel inventions. They serve as a convenient starting point for beginners.

  rising⭐ 67k

A mix of perennial favorites like `donnemartin/system-design-primer` and new AI-native tools like `abhigyanpatwari/GitNexus`. The trend is the integration of AI into core developer workflows, such as code analysis running entirely client-side.

• abhigyanpatwari/GitNexus

  A code intelligence tool that creates a knowledge graph from a codebase entirely within the browser. Its key feature is being serverless, which appeals to users concerned with privacy and who want a zero-setup analysis tool.

  new⭐ 31k
• Fincept-Corporation/FinceptTerminal

  A modern, open-source financial terminal for market analytics. Its repeated appearance on the trending list suggests sustained interest and development in creating open alternatives to proprietary financial data platforms like Bloomberg Terminal.

  repeated⭐ 16k
• penpot/penpot

  An open-source design and prototyping tool positioned as an alternative to Figma. Its trendiness reflects the continuous demand in the design and development communities for open, self-hostable collaboration tools.

  new⭐ 46k
• siddharthvaddem/openscreen

  An open-source alternative to Screen Studio for creating product demos. This tool addresses a specific need for developers and marketers to create polished videos without watermarks or subscriptions, a common driver for open-source adoption.

  cooling⭐ 33k
• HKUDS/DeepTutor

  An AI agent-based personalized learning assistant. This project demonstrates the application of the agent paradigm to the field of education, moving beyond simple Q&A bots to more interactive, customized tutoring experiences.

  cooling⭐ 22k
• donnemartin/system-design-primer

  A classic, comprehensive guide to system design for technical interviews. Its recurring presence is not due to new features but its status as a canonical resource for engineers preparing for interviews, often spiking during hiring seasons.

  new⭐ 345k
• GyulyVGC/sniffnet

  A network traffic monitoring tool written in Rust. It's trending due to the combination of a user-friendly interface for a complex task and its implementation in Rust, a language popular for building performant and secure system utilities.

  new⭐ 36k

These projects focus on building the foundational frameworks for creating and controlling AI agents. Repos like `coleam00/Archon` and `Tracer-Cloud/opensre` aim to make agent behavior more deterministic, reliable, and applicable to specific domains like SRE.

• NousResearch/hermes-agent

  An agent framework from a known AI research group. Its persistence on monthly charts indicates sustained interest from researchers and developers in exploring different agent architectures and frameworks beyond the most hyped platforms.

  cooling⭐ 120k
• lsdefine/GenericAgent

  This project claims to be a self-evolving agent that grows a skill tree autonomously. It attracts attention due to its ambitious goal of agent self-improvement, a key area of research in AGI, though such claims often require scrutiny.

  repeated⭐ 7.6k
• thunderbird/thunderbolt

  An AI framework emphasizing user control, model choice, and data ownership. This project taps into the growing developer sentiment against vendor lock-in and a desire for more transparent and controllable AI infrastructure.

  new⭐ 4.2k
• coleam00/Archon

  An open-source harness builder designed to make AI coding deterministic. This tool addresses a fundamental challenge of using LLMs in software engineering: their inherent randomness. It aims to make agent-based coding repeatable and testable.

  cooling⭐ 20k
• TauricResearch/TradingAgents

  A framework for using multiple LLM agents for financial trading. This is a specific, high-value application of the broader multi-agent system trend, showing how these concepts are being tested in quantitative finance.

  new⭐ 53k
• Tracer-Cloud/opensre

  An open-source toolkit for building AI-powered Site Reliability Engineering (SRE) agents. It applies the agentic AI trend to the complex domain of system operations, aiming to automate tasks like monitoring, incident response, and analysis.

  new⭐ 3.5k

This category is fragmented, featuring on-device showcases from Google, a voice AI project from Microsoft, and various application-specific tools. `Anil-matcha/Open-Generative-AI` is a list of models, not a tool, indicating high interest but also noise.

• microsoft/VibeVoice

  An open-source voice AI project from Microsoft. The backing of a major tech corporation indicates a serious strategic investment in open, multimodal AI capabilities, likely aimed at fostering an ecosystem around their technologies.

  repeated⭐ 42k
• Anil-matcha/Open-Generative-AI

  This is not a software tool but a large list aggregating over 200 generative AI models. Its popularity highlights the community's need for discovery and navigation in a crowded field, but the repository itself is low on technical substance.

  new⭐ 9.2k
• google-ai-edge/gallery

  A gallery of on-device machine learning examples from Google. This is primarily an educational resource and a developer marketing tool designed to showcase the capabilities of Google's on-device AI frameworks and encourage their adoption.

  cooling⭐ 22k
• microsoft/markitdown

  A simple Python utility from Microsoft for converting various file formats to Markdown. While a straightforward tool, its origin from a major company likely boosted its visibility. It addresses a common, practical need for developers and writers.

  cooling⭐ 118k
• google-ai-edge/LiteRT-LM

  A project from Google's on-device AI team, likely a lightweight language model runtime, but the repository lacks a description. Its trending status is based on its origin and name alone, indicating speculative interest in Google's on-device AI strategy.

  cooling⭐ 4.4k
• hugohe3/ppt-master

  An AI tool that generates editable PowerPoint (PPTX) files from documents. It's notable because it creates real, editable shapes and text boxes, not just images of slides. This is a practical application of AI for a widespread business task.

  cooling⭐ 8.6k

This category directly addresses the primary bottleneck for AI coding agents: limited context windows. Projects from infrastructure players (`zilliztech/claude-context`) and utilities (`mksglu/context-mode`) are tackling this with specialized code search and context compression.

• zilliztech/claude-context

  From the company behind the Milvus vector database, this tool provides code search to expand an agent's context. It's a key piece of infrastructure that solves the limited context window problem by integrating vector search directly into the agent workflow.

  new⭐ 9.8k
• gastownhall/beads

  Marketed as a 'memory upgrade for your coding agent'. This project is another take on solving the context and memory problem for agents, highlighting that this is a major area of active development and experimentation in the ecosystem.

  new⭐ 22k
• mksglu/context-mode

  A tool focused on optimizing the context window for AI agents across 14 different platforms. This is a pragmatic utility that addresses both the performance and cost issues associated with large context windows, making it widely applicable.

  new⭐ 11k
• HKUDS/RAG-Anything

  A general-purpose framework for Retrieval-Augmented Generation (RAG). Its broad name suggests it's a foundational library aiming to simplify the process of building RAG applications for any type of data, not just text.

  new⭐ 19k
• thedotmack/claude-mem

  A plugin specifically for Claude Code that captures and compresses coding session activity to inject as future context. This represents a tightly-integrated approach to the memory problem, tailored to a single agent platform.

  cooling⭐ 68k

This category features foundational model releases. The presence of a new large model from `deepseek-ai` and a specialized time-series model from Google (`timesfm`) shows continued parallel development in both general-purpose and domain-specific AI.

• shiyu-coder/Kronos

  A foundation model specifically trained on financial market data. Its existence points to a trend of developing domain-specific models that can potentially outperform general-purpose models on specialized tasks like financial forecasting.

  cooling⭐ 22k
• deepseek-ai/DeepSeek-V3

  A placeholder repository for a new model from DeepSeek AI. It gained stars based on an announcement alone, a common pattern where the hype for a new model precedes the actual release of any code or detailed information.

  new⭐ 103k
• google-research/timesfm

  A time-series foundation model from Google Research. This is a specialized model for a very common data science task (forecasting). Its presence shows that major labs are still investing in task-specific architectures, not just general LLMs.

  cooling⭐ 19k