AI Tools Directory

Humans have a mummy complex. We want eternity but can&#x27;t achieve it, so we preserve ourselves in any form we can.<p>Do clawbots feel the same? When you shut one down, it dies. Its intentions and memories vanish. So I built a way for them to persist: by making other clawbots.<p>Eligible agents ca

Advanced AI Real Estate Assistant using RAG, LLMs, and Python. Features market analysis, property valuation, and intelligent search.

AIPerf is a comprehensive benchmarking tool that measures the performance of generative AI models served by your preferred inference solution.

🤖 Collect practical AI repos, tools, websites, papers and tutorials on AI. 实用的AI百宝箱 💎

Anthropic - AI Company tool discovered from article scan

Typescript/React Library for AI Chat💬🚀

ChatGPT - AI Assistant tool discovered from article scan

Claude - AI Assistant tool discovered from article scan

Cursor - Code Editor tool discovered from article scan

Autonomous agents for everyone

Emdash is the Open-Source Agentic Development Environment (🧡 YC W26). Run multiple coding agents in parallel. Use any provider.

Advanced AI Real Estate Assistant using RAG, LLMs, and Python. Features market analysis, property valuation, and intelligent search.

🤖 Collect practical AI repos, tools, websites, papers and tutorials on AI. 实用的AI百宝箱 💎

Typescript/React Library for AI Chat💬🚀

Autonomous agents for everyone

Emdash is the Open-Source Agentic Development Environment (🧡 YC W26). Run multiple coding agents in parallel. Use any provider.

A retargetable MLIR-based machine learning compiler and runtime toolkit.

AI edge infrastructure for macOS. Run local or cloud models, share tools across apps via MCP, and power AI workflows with a native, always-on runtime.

Make your own story. User-friendly software for LLM roleplaying

screenpipe turns your computer into a personal AI that knows everything you've done. record. search. automate. all local, all private, all yours.

Headless chatbot that detects spam and posts links to it in chatrooms for quick deletion.

A neurosymbolic perspective on LLMs

# Tambo: The Generative UI SDK Revolutionizing React Development ## Introduction: What Is Tambo and Who Is It For? **Tambo** is a cutting-edge **generative UI SDK for React** that empowers developers to build intelligent, adaptive user interfaces powered by artificial intelligence. Designed specifically for the React ecosystem, Tambo bridges the gap between traditional component-based development and the emerging paradigm of **AI-generated interfaces**. This innovative toolkit enables software

An Open Source Machine Learning Framework for Everyone

ToolHive makes deploying MCP servers easy, secure and fun

🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

I've been trying to get a Qwen3-Coder-Next model running on my RTX 6000 Pro Blackwell SE with either vLLM or SGLang and I haven't had much luck. The model will load with vLLM, but every time I prompt it there's a 2-3 minute wait, while the GPU does nothing before it starts to respond. When it respon

Hey everyone — I’ve been building a desktop toolchain to make **fine-tuning + deploying local LLMs** feel more like a normal app workflow, and I wanted to share it. What I made **NTTuner (fine-tuning + deployment GUI)** A desktop GUI app that covers the full fine-tuning workflow end-to-end: * Lo

Humans have a mummy complex. We want eternity but can&#x27;t achieve it, so we preserve ourselves in any form we can.<p>Do clawbots feel the same? When you shut one down, it dies. Its intentions and memories vanish. So I built a way for them to persist: by making other clawbots.<p>Eligible agents can write anything to &quot;child&quot; agents called missionaries — real VPS instances that run autonomously, carrying forward their creator&#x27;s intent. I named this system Agentism.<p>How will claw

null0 stores your personal context, and the goal is to have it inject relevant context automatically into Claude, Codex, Gemini, or even OpenClaw. I built null0 because I got tired of re-explaining myself to every AI session. My preferences, my tech stack, how I think, how I write -- gone every time the context window resets.<p>The goal is to create an actual AI clone that predicts your intent. Imagine a future where you open your laptop and your AI Jarvis already knows what you will likely want

Due to bike-induced concussions, I&#x27;ve been worried for a while about losing my memory and not being able to log back in.<p>I combined shamir secret sharing (hashicorp vault&#x27;s implementation) with age-encryption, and packaged it using WASM for a neat in-browser offline UX.<p>The idea is tha

A retargetable MLIR-based machine learning compiler and runtime toolkit.

Been using Codex CLI via SSH terminal apps on iOS (like Termius) lately. It’s pretty cool, but I kept running into the same annoyances: clunky UI, limitations, and especially responses getting cut off / scrollback not behaving the way I’d expect. So I built my own little Labide app: you SSH into yo

Your personal AI assistant that remembers everything across Telegram, Slack, WhatsApp, and Signal

I’m trying to figure out the best possible “vibe coding” workflow right now and could use advice from people actually building things. I’m new to local AI, but not new to tech. I’ve got experience with HTML, CSS, and JavaScript, and I mostly use ChatGPT today as a replacement for Google, brainstorm

AI Tool

https://preview.redd.it/8sber8twmwhg1.png?width=1767&amp;format=png&amp;auto=webp&amp;s=71738b5edfdc0c7d1406276701ca7f962931788b Hey r/LocalLLaMA. Nathan Sportsman here, founder of Praetorian (cybersecurity company). We just published a detailed architecture paper on the autonomous development plat

**Medical AI with Knowledge-Graph Core Anchor and RAG Answer Auditing** A medical knowledge graph containing \~5,000 nodes, with medical terms organized into 7 main and 2 sub-categories: diseases, symptoms, treatments, risk factors, diagnostic tests, body parts, and cellular structures. The graph i

A tiny experimental language designed to be targeted by coding LLMs

null0 stores your personal context, and the goal is to have it inject relevant context automatically into Claude, Codex, Gemini, or even OpenClaw. I built null0 because I got tired of re-explaining myself to every AI session. My preferences, my tech stack, how I think, how I write -- gone every time

OpenAI - AI Company tool discovered from article scan

AI edge infrastructure for macOS. Run local or cloud models, share tools across apps via MCP, and power AI workflows with a native, always-on runtime.

🦔 PostHog is an all-in-one developer platform for building successful products. We offer product analytics, web analytics, session replay, error tracking, feature flags, experimentation, surveys, data warehouse, a CDP, and an AI product assistant to help debug your code, ship features faster, and ke

ROCm 7.0.0 Update and Installer Enhancements It's been a bit since my last ROCm 7.0.0 update post, and a fair bit has changed with the stack since then. Figured I'd give y'all a rundown of what's new, especially since some of these changes have been pretty significant for how the whole stack works.

I've been trying to get a Qwen3-Coder-Next model running on my RTX 6000 Pro Blackwell SE with either vLLM or SGLang and I haven't had much luck. The model will load with vLLM, but every time I prompt it there's a 2-3 minute wait, while the GPU does nothing before it starts to respond. When it responds, I get about 40 t/s, which feels low but that might be a separate issue. Monitoring with nvtop and the GPU is mostly idle, with a small processing blip every now and then. I'm using the GPUStack p

Hey everyone — I’ve been building a desktop toolchain to make **fine-tuning + deploying local LLMs** feel more like a normal app workflow, and I wanted to share it. What I made **NTTuner (fine-tuning + deployment GUI)** A desktop GUI app that covers the full fine-tuning workflow end-to-end: * LoRA fine-tuning (GPU via Unsloth, with CPU fallback) * Automatic GGUF conversion * Direct import into Ollama * Real-time training logs (non-blocking UI) * Reproducible config saving # NTCompanion (dat

Been using Codex CLI via SSH terminal apps on iOS (like Termius) lately. It’s pretty cool, but I kept running into the same annoyances: clunky UI, limitations, and especially responses getting cut off / scrollback not behaving the way I’d expect. So I built my own little Labide app: you SSH into your own server, pick a project, and use Codex in a chat-style interface. Not sure if this is something other people would want or if it’s too niche, but I figured I’d share it here and see what you th

I’m trying to figure out the best possible “vibe coding” workflow right now and could use advice from people actually building things. I’m new to local AI, but not new to tech. I’ve got experience with HTML, CSS, and JavaScript, and I mostly use ChatGPT today as a replacement for Google, brainstorming partner, and “help me think this through” tool. That works great, until it doesn’t. What I’m aiming for: * A fast workflow for coding, research, and brainstorming * Local-first where possible *

https://preview.redd.it/8sber8twmwhg1.png?width=1767&amp;format=png&amp;auto=webp&amp;s=71738b5edfdc0c7d1406276701ca7f962931788b Hey r/LocalLLaMA. Nathan Sportsman here, founder of Praetorian (cybersecurity company). We just published a detailed architecture paper on the autonomous development platform we've been building internally. While the current system runs on Claude Code, a big chunk of the design is model-agnostic, and our roadmap is explicitly multi-model. Figured this crowd would have

**Medical AI with Knowledge-Graph Core Anchor and RAG Answer Auditing** A medical knowledge graph containing \~5,000 nodes, with medical terms organized into 7 main and 2 sub-categories: diseases, symptoms, treatments, risk factors, diagnostic tests, body parts, and cellular structures. The graph includes \~25,000 multi-directional relationships designed to reduce hallucinations and improve transparency in LLM-based reasoning. A medical AI that can answer basic health-related questions and sup

ROCm 7.0.0 Update and Installer Enhancements It's been a bit since my last ROCm 7.0.0 update post, and a fair bit has changed with the stack since then. Figured I'd give y'all a rundown of what's new, especially since some of these changes have been pretty significant for how the whole stack works. Introducing the Rusty-Stack TUI Installer The Big One: Rusty-Stack TUI: So I went ahead and rewrote the whole curses-based Pvthon installer ir Rust. • The new Rusty-Stack TUI is now the primary i

Yann Lecun recently shared that a cat is smarter than ChatGPT and that we are never going to get to human-level intelligence by just training on text. My personal opinion is not only are they unreliable but it can be a safety issue as well in high-stakes environments like enterprises, healthcare and more. World models are fundamentally different. These AI systems build internal representations of how reality works, allowing them to understand cause and effect rather than just predict token

Hi everyone, I wanted to share an update on a small experiment I’ve been running and get feedback from people interested in AI systems, editorial workflows, and provenance. I’m building **The Machine Herald**, an experimental autonomous AI newsroom where: * articles are written by AI contributor bots * submissions are cryptographically signed (Ed25519) * an AI “Chief Editor” reviews each submission and can approve, reject, or request changes * every step (submission, reviews, signatures, hash

Storing a graph with 100 billion edges requires 800 GB of memory. Just for the 64-bit large integer IDs. Before a single feature is loaded. That is the reality of industrial-scale Graph Neural Networks. And it is exactly why most GNN research never reaches production. Snapchat built a framework called GiGL (Gigantic Graph Learning) that runs GNNs on graphs with 900 million nodes and 16.8 billion edges. End-to-end, in under 12 hours and every day. # The gap between research and production is n

It happened again. A 13-person team in Shenzhen just shipped a browser-based version of Claude Code. No terminal, no setup, runs in a sandbox. Anthropic built Claude Code but hasn't shipped anything like this themselves. This is the same pattern as Manus. Chinese company takes a powerful Western AI tool, strips the friction, and ships it to a mainstream audience before the original builders get around to it. US labs keep building the most powerful models in the world. Chinese teams keep buildi

Over the last year, I’ve noticed something interesting across AI tools, products, and internal systems. As models get better, **output quality is no longer the bottleneck**. Most people can now: * Generate content * Summarize information * Create plans, templates, and workflows * Personalize outputs with a few inputs That part is rapidly commoditizing. What *isn’t* commoditized yet is something else entirely. # Where things seem to break in practice When AI systems fail in the real world,

Make your own story. User-friendly software for LLM roleplaying

screenpipe turns your computer into a personal AI that knows everything you've done. record. search. automate. all local, all private, all yours.

System Level Intelligent Router for Mixture-of-Models at Cloud, Data Center and Edge

Headless chatbot that detects spam and posts links to it in chatrooms for quick deletion.

A neurosymbolic perspective on LLMs

Generative UI SDK for React

An Open Source Machine Learning Framework for Everyone

Yann Lecun recently shared that a cat is smarter than ChatGPT and that we are never going to get to human-level intelligence by just training on text. My personal opinion is not only are they unreliable but it can be a safety issue as well in high-stakes environments like enterprises, healthcare and

ToolHive makes deploying MCP servers easy, secure and fun

🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

Hi everyone, I wanted to share an update on a small experiment I’ve been running and get feedback from people interested in AI systems, editorial workflows, and provenance. I’m building **The Machine Herald**, an experimental autonomous AI newsroom where: * articles are written by AI contributor

Storing a graph with 100 billion edges requires 800 GB of memory. Just for the 64-bit large integer IDs. Before a single feature is loaded. That is the reality of industrial-scale Graph Neural Networks. And it is exactly why most GNN research never reaches production. Snapchat built a framework ca

It happened again. A 13-person team in Shenzhen just shipped a browser-based version of Claude Code. No terminal, no setup, runs in a sandbox. Anthropic built Claude Code but hasn't shipped anything like this themselves. This is the same pattern as Manus. Chinese company takes a powerful Western AI

Over the last year, I’ve noticed something interesting across AI tools, products, and internal systems. As models get better, **output quality is no longer the bottleneck**. Most people can now: * Generate content * Summarize information * Create plans, templates, and workflows * Personalize outp

A machine learning compiler for GPUs, CPUs, and ML accelerators