Denoise

Anthropic's own entry in financial services signals a strategic move by foundational model providers to build vertical-specific agent examples, moving up the application stack.

• anthropics/financial-services

  A repository from Anthropic demonstrating financial services use cases. Its presence is a signal that foundation model providers are moving up the stack to build vertical-specific examples, directly engaging with high-value agent applications.

  rising⭐ 14k

This space is about commoditizing access to powerful coding models. Tools like decolua/9router and Alishahryar1/free-claude-code focus on abstracting APIs and routing to free endpoints, while warpdotdev/warp integrates agents directly into the terminal workflow.

• Hmbown/DeepSeek-TUI

  A terminal-based coding agent specifically for DeepSeek models. This points to the specialization of agent tooling around particular open-source models, not just proprietary ones like Claude or GPT.

  new⭐ 21k
• Alishahryar1/free-claude-code

  A tool that provides free access to Claude Code through various interfaces. This is part of a cat-and-mouse trend where developers create wrappers and proxies to use powerful, often restricted models without direct API costs.

  cooling⭐ 23k
• decolua/9router

  A router that connects various AI coding tools to free model endpoints, with features like auto-fallback and token optimization. It represents the commoditization of model access, abstracting away the choice of provider and focusing on availability and cost.

  rising⭐ 5.2k
• warpdotdev/warp

  A terminal reimagined as an 'agentic development environment'. This is a significant product direction, integrating agent capabilities directly into a core developer tool, rather than having them as separate scripts or web interfaces.

  new⭐ 57k

The focus has decisively shifted from raw model access to agent orchestration and management. Projects like ruvnet/ruflo, multica-ai/multica, and openai/symphony are all building the 'Kubernetes for agents', indicating a maturing ecosystem.

• ruvnet/ruflo

  Presents itself as an enterprise-grade orchestration platform for Claude agents. Its emergence indicates a move towards professionalizing agent development, providing infrastructure for managing complex, multi-agent workflows beyond simple scripts.

  new⭐ 47k
• NousResearch/hermes-agent

  An agent from Nous Research, a well-known group in the open-source AI space. While the description is minimal, its continued presence suggests ongoing interest in developing general-purpose, self-improving agents.

  cooling⭐ 139k
• TauricResearch/TradingAgents

  A framework for building multi-agent systems for financial trading. This is a recurring theme today, highlighting finance as a key early-adopter vertical for complex agentic workflows and automated decision-making.

  repeated⭐ 72k
• multica-ai/multica

  An open-source platform to manage coding agents as team members. It aims to solve the operational challenges of using agents at scale, such as task assignment and skill compounding, part of the broader trend of building agent management systems.

  cooling⭐ 26k
• browserbase/skills

  An SDK for Claude agents that includes a web browsing tool. This provides a key capability for agents to interact with the live web, grounding their operations in real-world data beyond their training set. It's a foundational component for autonomous research agents.

  new⭐ 2.6k
• awslabs/aidlc-workflows

  A workflow steering system for AI coding agents from AWS Labs. This shows that major cloud providers are investing in the control and management layer for agents, aiming to make their behavior more predictable and reliable in a software development lifecycle.

  new⭐ 1.6k
• HKUDS/AI-Trader

  Another fully-automated, agent-based trading system. Its presence alongside similar projects reinforces the concentration of effort in applying agent technology to financial markets, likely driven by the clear potential for automation.

  new⭐ 14k
• 1jehuang/jcode

  Described as a 'Coding Agent Harness'. This term is becoming common, referring to frameworks that structure and manage the execution of coding agents. It represents a piece of the emerging standard toolkit for agent development.

  new⭐ 5.0k
• lsdefine/GenericAgent

  A self-evolving agent that grows a skill tree from a small seed. This project explores the concept of agent autonomy and self-improvement, a more research-oriented goal compared to the more pragmatic orchestration tools also trending.

  cooling⭐ 9.7k
• lobehub/lobehub

  Positions itself as a collaborative platform for working with 'agent teammates'. The focus on multi-agent collaboration and skill growth suggests an ambition to build a 'GitHub for agents', a central hub for agent interaction and development.

  new⭐ 76k
• cocoindex-io/cocoindex

  An incremental engine designed for agents operating over long time horizons. This tackles a key technical challenge: maintaining state and making progress on complex, multi-step tasks without getting lost or repeating work.

  new⭐ 9.1k
• coleam00/Archon

  An open-source 'harness builder' for AI coding. The goal is to make AI-driven coding more deterministic and repeatable, addressing the reliability issues that often plague agent-based software development. Part of the broader 'agent harness' trend.

  cooling⭐ 21k
• openai/symphony

  An OpenAI project for managing development work via autonomous agent runs. As a tool from a leading lab, it validates the industry-wide shift towards using agentic workflows to automate and manage software implementation tasks.

  new⭐ 23k

This category solidifies the idea of a 'skill' as a new software artifact. These are not traditional code libraries, but collections of structured prompts for agents, authored by prominent developers like Addy Osmani and Matt Pocock.

• forrestchang/andrej-karpathy-skills

  A single markdown file of prompts to improve Claude's coding behavior, based on Andrej Karpathy's observations. Its popularity highlights the emergence of prompt engineering collections, or 'skills', as a shareable and valuable asset.

  cooling⭐ 120k
• addyosmani/agent-skills

  A library of production-grade skills for coding agents from a well-known Google engineer. This is a prime example of the 'skill' abstraction layer, providing reusable, high-quality prompts to make agents more effective at common engineering tasks.

  repeated⭐ 34k
• mattpocock/skills

  A personal collection of 'skills' for Claude from a prominent developer in the TypeScript community. This, along with similar repos, shows a pattern of expert developers codifying their interaction patterns with LLMs into shareable assets.

  repeated⭐ 66k
• flutter/skills

  An empty repository from the official Flutter team. Its mere existence under this name is a strong signal that major development frameworks are exploring standardized agent skills as a way to improve developer experience and automation.

  new⭐ 1.5k
• ComposioHQ/awesome-codex-skills

  A curated list of skills for Codex, an older but still relevant model. This shows that the concept of skills isn't limited to the newest models and that there's a need to catalogue and share effective automation recipes.

  cooling⭐ 7.6k

This category contains perennial tools with consistent interest. soxoj/maigret for OSINT and Z4nzu/hackingtool as a script collection represent established use cases rather than a new structural trend.

• soxoj/maigret

  An OSINT tool for collecting information on individuals by username across thousands of sites. This is a long-standing, popular project in the security space that trends periodically due to its broad utility.

  repeated⭐ 26k
• Z4nzu/hackingtool

  A collection of hacking scripts and tools. These kinds of all-in-one repositories are common and attract consistent attention from security hobbyists and beginners, though they rarely represent new technical developments.

  cooling⭐ 73k

This category focuses on AI-driven content generation beyond code and text. Projects for automated video (AIDC-AI/Pixelle-Video), music UI (fspecii/ace-step-ui), and native presentations (hugohe3/ppt-master) show multimodal AI being productized.

• AIDC-AI/Pixelle-Video

  An AI-powered engine for fully automated short video creation. This represents the application of generative AI to content marketing and social media, automating a time-consuming creative process.

  repeated⭐ 14k
• OpenBMB/VoxCPM

  A multilingual, tokenizer-free text-to-speech model. The 'tokenizer-free' aspect is technically significant, potentially allowing for better handling of diverse languages and vocal nuances. From a notable Chinese AI lab.

  cooling⭐ 18k
• microsoft/markitdown

  A utility from Microsoft for converting various document formats to Markdown. A practical tool for developers and technical writers looking to standardize documentation into a plain-text, version-controllable format.

  cooling⭐ 122k
• fspecii/ace-step-ui

  A user interface for an open-source AI music generation model, positioned as an alternative to Suno. This highlights a common pattern where the community builds user-friendly frontends for powerful but hard-to-use backend models.

  new⭐ 3.3k
• hugohe3/ppt-master

  An AI tool that generates editable PowerPoint (PPTX) files from documents. Unlike tools that create image-based slides, this one generates native shapes and animations, making the output far more useful for business users.

  cooling⭐ 13k

This category shows activity at the lower level of the stack. z-lab/dflash offers a specific performance optimization for inference, while shiyu-coder/Kronos reinforces the pattern of developing foundation models for the finance vertical.

• z-lab/dflash

  An implementation of a speculative decoding technique to accelerate diffusion models. This is a low-level infrastructure project focused on improving the performance and efficiency of model inference, a constant area of research.

  new⭐ 3.7k
• shiyu-coder/Kronos

  Described as a foundation model for financial markets. This is another data point showing finance as a key area for developing specialized, domain-specific models, moving beyond general-purpose LLMs.

  cooling⭐ 24k

These projects are building the persistence and memory layer for agents. Solutions range from specialized research agents like virattt/dexter to tightly-integrated Claude plugins like thedotmack/claude-mem, showing a need for long-term context management.

• virattt/dexter

  An autonomous agent for deep financial research. It combines agentic control flow with data retrieval to automate a complex, knowledge-intensive task. It fits into both the agent and finance themes of the day.

  new⭐ 25k
• LearningCircuit/local-deep-research

  A framework for conducting research using local models and local documents. The emphasis on privacy and local execution addresses a key enterprise concern and shows the maturing capabilities of smaller models running on consumer hardware.

  new⭐ 6.6k
• thedotmack/claude-mem

  A plugin for Claude Code that provides long-term memory by capturing and compressing session context. This is a practical tool addressing a fundamental limitation of LLMs (limited context windows) and is specific to the Claude ecosystem.

  cooling⭐ 74k

A practical but varied category. CloakHQ/CloakBrowser addresses a specific pain point in web automation (bot detection), while docusealco/docuseal is a classic open-source alternative play. The tutorial datawhalechina/hello-agents shows demand for agent-related education.

• CloakHQ/CloakBrowser

  A modified version of Chromium designed to evade bot detection. This is a specialized developer tool for web scraping and automation, addressing the increasing sophistication of anti-automation measures on the web.

  new⭐ 2.6k
• docusealco/docuseal

  An open-source alternative to DocuSign. This project follows a classic open-source playbook: providing a self-hostable, transparent alternative to a popular proprietary SaaS product. It appeals to users concerned with data ownership and cost.

  new⭐ 16k
• datawhalechina/hello-agents

  A Chinese-language tutorial on building AI agents from scratch. Its popularity indicates a strong demand for educational resources as developers rush to acquire skills in this rapidly emerging paradigm of agent-based software.

  new⭐ 44k