📦 kavinsood/yaos

YAOS: Real-Time Obsidian Sync Without the Terminal

YAOS makes Obsidian notes appear on every device the instant you type them, without needing a terminal, a database, or a paid subscription. It runs on a free Cloudflare Worker that you deploy with a single click, turning your personal infrastructure into a live sync engine. The standout feature is its use of CRDTs to merge edits in real time, so two people typing in the same note never get stuck with conflicting copies. It works entirely locally, keeping your vault as plain Markdown files while syncing changes instantly across desktop and mobile, even offline, and integrates cleanly with scripts or AI agents that edit files directly on disk.

🆔 @hackernewsgithubprojects

📦 docwriter-org/plain-writing-skill

Plain Writing Skill: Fixing AI's Wordy Output

Plain Writing Skill is the instruction set that forces AI to cut the corporate fluff and speak like a human. It acts as a strict editor for your chatbot, stripping away jargon, empty buzzwords, and confusing sentence structures from its responses. Instead of letting an AI bury the good news in a paragraph of marketing-speak, this tool ensures the main point is delivered first and clearly. It is essentially a quality control gate that makes sure your code assistant actually explains its work in simple, everyday language. It is a clever way to keep your AI honest and readable.

🆔 @hackernewsgithubprojects

📦 rhoban/microban

Microban: The Robot You Can Print at Home

Microban is a thirty centimeter tall humanoid robot that you can print entirely on your desktop 3D printer. For around six hundred dollars, this open source creation gives you a fully functional body with nineteen servo motors and a tiny computer brain. It is built specifically for students and hobbyists who want to understand how bipedal walking works without buying expensive commercial hardware. Because the design is free to modify, you can experiment with movement algorithms, mechanical tweaks, and software upgrades directly on the physical unit. The goal is to make advanced robotics research accessible to anyone with a standard printer and basic tools.

🆔 @hackernewsgithubprojects

📦 drhazemali/enterprise-system-design

The Blueprint for Enterprise AI

Enterprise System Design is the architecture course that finally treats AI like production infrastructure. It stops pretending that building a system is just drawing boxes and arrows. Instead, it walks you through the messy reality of how data moves, how things break, and how you keep them safe. You will learn to handle partial failures and strict security without freezing. It covers everything from basic cloud logic to complex model serving. This is not a theory class. It is a practical map for engineers who need to build things that actually work under pressure. If you want to design systems that survive real traffic, this is your starting point.

🆔 @hackernewsgithubprojects

📦 vibeinging/deepseek-harness-desktop-app

The Local AI Workbench

DeepSeek Harness Desktop App lets you inspect every single decision an AI makes before it touches your files. Most coding tools just show the final code, but this application displays the full reasoning trail, showing exactly which tools ran and what data the model accessed to reach its answer. It runs entirely on your machine, keeping your projects, files, and chat history private instead of sending them to a cloud server. The app creates safe, isolated spaces for AI to experiment with your code without breaking your main project. This gives you total control and transparency when using artificial intelligence for your daily software development tasks.

🆔 @hackernewsgithubprojects

📦 letterstory/lettertrace

Lettertrace: Track Your Brand in AI Answers

Lettertrace lets you see exactly how your brand shows up when people ask ChatGPT or Claude for recommendations. Instead of guessing, it sends realistic questions to these AI models using your own API keys and tracks whether your name is mentioned, recommended, or ignored. It even compares your visibility against competitors over time. Since it is open source and self-hosted, you keep full control of the data. If you want to know if AI assistants actually talk about your business, this tool gives you the honest answer.

📰 https://news.ycombinator.com/item?id=49299569

🆔 @hackernewsgithubprojects

📦 1038lab/comfyui-qwentts

ComfyUI-QwenTTS: Design Any Voice With Words

A new tool lets you type a simple sentence and get a completely custom human voice back. ComfyUI-QwenTTS brings powerful speech synthesis into your local node-based workflow, allowing you to generate audio that sounds exactly how you described it. This project solves the problem of rigid voice options by letting you create unique speaker profiles using plain English descriptions, rather than choosing from a limited menu. It works locally on your machine, keeping your data private while giving you total creative control. Whether you need a narrator for a video or a specific character voice, this setup turns text into realistic sound without sending your files to the cloud.

🆔 @hackernewsgithubprojects

📦 anionex/dsh-vision-toolkit

DSH Vision Toolkit: Give Your Text AI Real Eyes

DSH Vision Toolkit is the plugin that gives text-only AI agents the ability to actually see and verify visual work. Instead of guessing at what a screenshot looks like, this tool lets an AI model measure pixel differences, read text from long images, and rebuild user interfaces from sketches or photos with precise accuracy. It’s perfect for developers who want their AI to handle visual debugging without constantly switching contexts. The coolest part is that it doesn’t just describe what it sees; it proves it by comparing rendered code against original designs down to the exact pixel. Now your AI can build and verify interfaces with human-like precision.

🆔 @hackernewsgithubprojects

📦 rbadillap/pastecn

The Code Paster That Feeds Your CLI

Paste any snippet of code and instantly get a link that works directly with the shadcn command line tool. This project solves the awkward friction of sharing custom UI components. Usually, you would have to copy your code into a text file, email it, or hope your friend manually recreates the file structure. Here, the system wraps your code in a permanent, immutable URL that is fully compatible with standard developer workflows. You can even share complex, multi-file setups with a single click, and your colleagues install the entire package in one step without any manual copying. It is the missing bridge between casual code sharing and professional component management.

🆔 @hackernewsgithubprojects

📦 naticha/bunwv

The Browser That Thinks Like Your AI Agent

Bunwv is the headless browser CLI that finally lets AI coding assistants drive a real web page without crashing. Unlike tools that vanish after every click, Bunwv keeps a persistent browser daemon alive, so your login session, form data, and page state survive across every command. This is a massive win for automation because it means your AI agent can navigate a site, fill out a form, and verify the result all in one continuous flow, just like a human would. It is built specifically for AI tools, speaking a clean JSON language that prevents errors and keeps the workflow efficient.

🆔 @hackernewsgithubprojects

📦 nilbuild/claude-run

Claude Run: See Your AI History

Open a browser tab and see every conversation you’ve had with Claude Code. The project called Claude Run takes the hidden text files where your chat history lives and turns them into a clean, readable interface. It solves the problem of lost context by letting you search past prompts, filter by project, and watch live updates as Claude responds. This is useful when you need to recall a specific solution or want to resume a previous task without digging through messy logs. The standout feature is the live stream view, which lets you monitor ongoing sessions in real time right from your desktop.

🆔 @hackernewsgithubprojects

📦 elie222/rakazo

Rakazo: Own Your AI Teammates

Rakazo is the open-source bot platform that gives every agent its own private computer to think on. While most tools just chat in text boxes, this project spins up a full Linux desktop for each bot, letting it browse the web, run code, and handle files visually right in front of you. You plug in your favorite AI model and choose where it runs, keeping total control over the sandbox. It is a powerful way to build autonomous digital coworkers that you actually own, not rent. Check out the repo to see how your next AI teammate works.

🆔 @hackernewsgithubprojects

📦 sheberdavid/v4-flash-godmode-opencode-go

V4 Flash Godmode: Fixing AI Speed

V4 Flash Godmode is the agent preset that fixes the sloppy coding habits of fast AI models. The problem is simple. When you use DeepSeek V4 Flash without the right instructions, it rushes into coding. It skips deep planning and delivers shallow, broken work. This tool, built for the opencode go provider, injects a specific persona that forces the model to slow down. It demands deep architectural thinking and self-testing before any code is written. In one real-world test of a four-stroke engine simulation, this preset increased the model's planning depth from thirty thousand characters to over three hundred thousand. The result was a bug-free, numerically verified deliverable.

🆔 @hackernewsgithubprojects

📦 serradura/okf-gem

OKF: Portable Knowledge for AI Agents

Give your coding agents a portable, local brain. OKF lets you package project knowledge into simple Markdown bundles that any tool can read and verify, solving the problem of fragmented notes that confuse AI assistants. Think of it as a standardized format for runbooks and decisions, complete with a built-in checker to catch errors before your agent sees them. It even renders an interactive graph so you can visualize how your knowledge connects, keeping everything offline and in your control. Next time you hand off a complex project, wrap your insights in this format so your AI teammates actually understand what to do.

🆔 @hackernewsgithubprojects

📦 gaossr/best-claude-hud

The Rust HUD for Claude Code

Best Claude Hud is the terminal companion that finally gives you a live dashboard for your coding sessions without cluttering your screen. If you use the Claude Code agent, you know how disorienting it can feel when you lose track of which model is running or how much context you’ve burned through. This tool solves that by injecting a sleek, real-time status bar directly into your terminal. It pulls live data straight from Claude’s official status line, showing you exactly which model is active, your current Git branch, and most importantly, how much of your context window is being used.

🆔 @hackernewsgithubprojects

📦 cristicretu/diri

Diri: The Dashboard for Parallel Coding Agents

Diri is the native desktop app that finally lets you run multiple AI coding agents at once without losing track of any of them. It creates separate workspaces for tools like Claude Code, Codex, and Cursor, so they can all edit different parts of your project in parallel without crashing into each other. The standout feature is its live status system, which actually reads the screen to tell you if an agent is still working, waiting for your input, or finished. You don’t have to babysit ten open terminals anymore.

🆔 @hackernewsgithubprojects

📦 pax-beehive/paxm

PAXM: Memory for Coding Agents

Coding agents have a nasty habit of forgetting everything the moment you close the chat window, forcing you to re-explain your project rules from scratch every single time. PAXM solves this by acting as a persistent, provider-neutral memory layer that works across tools like Codex, Claude Code, and OpenCode. It quietly saves your decisions and project context, then recalls them in future sessions so you never have to repeat yourself. The best part is that it starts locally with simple SQLite, so you get zero-setup benefits without needing API keys or expensive cloud services.

🆔 @hackernewsgithubprojects

📦 xai-org/grok-build-plugin-cc

Grok Build Plugin for Claude Code

XAI released a plugin that lets you hand off coding tasks to Grok while staying inside Claude Code. The bridge, called Grok Build, takes your current project state and sends it to Grok for deep reviews, harsh critiques, or full task delegation. It’s basically a second brain for your code, running in the background so you don't have to wait around. You can even import your entire Claude session into Grok to keep working where you left off. It’s a clever way to get two different AI models working together on the same codebase, giving you a fresh set of eyes without switching tools.

🆔 @hackernewsgithubprojects

📦 u-ichi/compact-plus

Compact Plus Fixes AI Memory Loss

Your coding agent forgets what it just built the moment you clear its context. Compact Plus fixes this by automatically saving a structured summary of your work right before compaction, then handing it back immediately after. It works invisibly in the background for Claude Code and Codex, ensuring your agent remembers its plan, decisions, and current task without you needing to do anything extra. This plugin is the safety net that keeps your AI focused, so your code never loses its place during long, complex sessions.

🆔 @hackernewsgithubprojects

📦 haddenhunter/skillforge

SkillForge: The npm for AI Agents

SkillForge turns AI coding agent capabilities into installable packages, finally giving your tools a proper home like npm. Instead of just pasting random prompt snippets, you can publish, version, and securely share specific skills across different AI models like Claude Code or Cursor. The standout feature is its strict sandboxing, which locks each skill down so it can only access the exact files and network hosts you allow, preventing your AI from accidentally reading sensitive data or causing chaos. It works locally first, so your code never leaves your machine, and it even lets you evaluate skills against baselines to ensure they perform consistently before you deploy them.

🆔 @hackernewsgithubprojects