← Overview
GitHub Trending · Thursday

5 November 2026

16 repos rising that day. Updated

metadataenterprise dataauto-ontologygraph layeragentMCP, CLI, APIdefinitionsreconciled
+2K that day★ 5.3KPython

Graph ontology layer so AI agents read data correctly

A Python library from NVIDIA that builds a context layer from enterprise metadata, reconciles definitions across systems, and exposes it to agents through MCP, CLI, and API.

  • Generates ontologies from metadata and connects existing ones.
  • Supports deterministic query generation and more explainable context for LLM and agent workflows.
worktrunkgit repoworktree Aagent 1worktree Bagent 2worktree Cagent 3
+531 that day★ 11.1KRust

Git worktree CLI for parallel AI agent workflows

Worktrunk manages Git worktrees from the command line, designed for running several AI agents in parallel on one repo. Written in Rust.

  • The README says Worktrunk was released earlier this year and quickly became the most popular git worktree manager.
  • Docs live at worktrunk.dev, and the license is MIT.
skillsmodularany modelagent runs themyour versionhack, adapt
+1.4K that day★ 313KShell

Small, composable agent skills from the author's .agents folder

A set of agent skills the author uses daily for real engineering, not vibe coding. Designed to be small, easy to adapt, and usable with any model.

  • The README contrasts it with GSD, BMAD, and Spec-Kit, which take ownership of the process.
  • Skills are composable so you can reshape them to your own needs.
SDK or CLIAWS, Terraformflociport 4566AWS serviceslocal emulationdockercompose up
+1.1K that day★ 29.7KJava

Free local AWS emulator with no account needed

Floci runs AWS-shaped services on your own machine for development, testing, and CI, with no account, auth token, or paid feature gates. Positioned as a LocalStack alternative.

  • Point the AWS SDK, CLI, Terraform, CDK, OpenTofu, or your test suite at http://localhost:4566.
  • Getting started takes just docker compose up.
trydocker compose up
sessiontool callsclaude-memcapture, compressmemorysummariesnext sessioncontext injected
+1.2K that day★ 113KTypeScript

Cross-session memory for AI coding agents

Records what an agent does during a session, compresses it with AI, and injects relevant context into later sessions. Works with Claude Code, Codex, Gemini, OpenCode, and more.

  • Captures tool usage observations, generates semantic summaries, and makes them available to future sessions.
  • Works with many agents, including OpenClaw, Hermes, and Copilot.
trynpx claude-mem install
desktop appaccessibility treeagent-desktopJSON with refsagentpicks a refclick or typethen re-read
+2.5K that day★ 15.3KRust

Computer use CLI for AI agents via the accessibility tree

agent-desktop reads a desktop app's accessibility tree as JSON with a ref for every element, then the agent clicks or types on a ref. No screenshots and no model inside.

  • Runs on macOS, Windows, and Linux, shipped as a CLI and a C library.
  • The agent re-reads the UI to check the result and repeats; the loop lives in the agent, not in agent-desktop.
  • The README warns that snapshots, clipboard reads, and screenshots can contain private desktop content.
specificationstart hereplanningwith templatesimplementationby the agentconvergencedocumented
+808 that day★ 147KPython

Structured process toolkit for coding agents

Spec Kit gives coding agents structured processes, reusable templates, and documented outcomes. Use a process as is, customize it, or bring your own.

  • Spec-Driven Development carries a specification through planning, implementation, and convergence.
  • There are also processes for fixing broken behavior and assessing an idea.
pull requestcode changesPR AgentAI reviewfeedbackreview comments
+280 that day★ 14.8KPython

Open-source AI agent that reviews pull requests

PR-Agent is an AI-powered code review agent and a community-maintained legacy project of Qodo. This repo is not Qodo's free offering for open-source projects.

  • The README stresses this repo is separate from Qodo's main code review product.
  • Community-maintained and supported by sponsors.
topicor keywordscriptAI modelfootagesubtitles, musicshort videoHD output
+617 that day★ 139KPython

Generate HD short videos automatically from one topic

MoneyPrinterTurbo takes a topic or keyword, writes the video script, matches footage, generates subtitles and background music, and composes an HD short video.

  • Available through a WebUI and an API.
  • Combines large AI models and an automated workflow in one tool.
gstack23 toolsCEOrethinks producteng managerarchitectureQA leadopens a browser
+290 that day★ 139KTypeScript

Garry Tan's Claude Code setup as a virtual engineering team

gstack is 23 slash-command, Markdown-based tools that turn Claude Code into a virtual engineering team with CEO, designer, eng manager, QA, and other roles. MIT license.

  • Roles include CEO, designer, eng manager, reviewer, QA, security officer, and release engineer.
  • QA opens a real browser; the security officer runs OWASP and STRIDE audits.
checkpoint1.56 TB on diskC99 engine176 KBtoken26.5 s at 8 GBRAM8.24 GB peak
+351 that day★ 10.4KC

Kimi K3 inference on one CPU in plain C99

Runs Kimi K3 (2.78 trillion parameters) on a single CPU with 8.24 GB peak RSS, using portable C99 with no BLAS, framework, or GPU. The whole engine is 176 KB.

  • The checkpoint is 1.56 TB on disk; on an 8 GB laptop the whole model streams from disk every step, about 26.5 s per token.
  • More RAM only buys speed, not a different answer.
imagelayers, layoutdisc dataphysical limitsburned discvisible image
+1.8K that day★ 6.7KC#

Burn visible images onto CD, DVD, and BD surfaces

Disc Image Studio converts an image into data that fits the physical limits of CD, DVD, and single-layer BD-R, so the burned disc surface shows the picture. Cross-platform.

  • Windows: download win-x64.exe. Linux and macOS use an experimental CLI.
  • The .NET runtime is bundled in the release packages.
try./disc-image-studio --help
90 daysDevOpsprinciplesfoundationsprocesseshow teams worktoolingtools to learnuse casesreal scenarios
+97 that day★ 30.5KShell

A 90-day learning map for DevOps

A learning-in-public project started on 1 January 2022 that is now a community learning map for DevOps: principles, processes, tooling, and use cases.

  • It began as a 90-day journal, 1 January to 31 March 2022.
  • Contributors are invited to add useful learning resources.
ScraplingPythonfetchersHTTP or browserparserselect elementsspidersconcurrent crawlproxiesrotation
+2K that day★ 96.7KPython

Web scraping framework from one request to full crawls

Scrapling is a Python web scraping framework with fast HTTP requests, stealth browsers, parsing, and concurrent crawling. It includes a CLI and MCP.

  • Main parts: selection methods, fetchers, spiders, and proxy rotation.
  • Covers everything from a single request to a full-scale crawl.
newsstoriesHorizonAI rankingbriefingEN and ZH
+121 that day★ 12.6KPython

AI news radar with daily briefings in two languages

Horizon generates daily news briefings in English and Chinese with AI help, built around a ranked list of stories. A live demo and a configuration guide are available.

  • The README shows a Ranked Daily Briefing plus context, summary, and discussion.
  • Uses uv tooling and has a demo site.
agentClaude, Cursorskillsmodel, priceScenario MCPmakes assetsDCC appsexpert teams
+224 that day★ 2.5KPython

Agent skills for making images, video, audio, and 3D

Skills that teach coding agents to create production-ready content through the Scenario MCP: images, video, audio, textures, skyboxes, 3D assets, and custom models. Picks the model and prices before spending.

  • Keeps characters and brands consistent across outputs.
  • Expert skill teams drive Blender, Maya, ZBrush, Unreal, and Unity.