Tools Bench.

Product launches and open-source repos with enough signal to earn a second look.

Last Brew Time: Aug 13, 2026, 11:03 AM PT

Insight

Builders here are quietly moving the agent stack off the cloud and onto your own machine

01.LOCAL-FIRST AI COMPUTING

02.REUSABLE SKILL LIBRARIES TEACHING AGENTS NEW TRICKS

03.OPEN ALTERNATIVES TO CLOSED INFRASTRUCTURE

Featured

GitHub168.9K

anthropics/skills

Public repository for Agent Skills

youtube
Don't Build Agents, Build Skills Instead – Barry Zhang & Mahesh Murag, Anthropic

Market Signal

Why It Has Market Pull

This official Agent Skills repository has become one of the most widely adopted AI developer resources of the year, with an ecosystem of third-party integrations spanning multiple coding agents and IDEs. Verified star count and active issue volume confirm massive, sustained community engagement rather than passive stargazing.

  • Roughly 169,000 GitHub stars and 20,000+ forks, still receiving commits 11 months after launch
  • Over 1,000 open issues and 700+ open pull requests reflect active real-world use, not a static template dump
  • Adopted well beyond its origin lab: major coding IDEs and agent tools have all built support for the same skill format
  • An active six-category discussion board hosts genuine third-party projects built on top of the format
  • Functions more as a foundational content/template standard than a standalone product, but its adoption footprint is unusually broad for that category

feedbacks

What People Are Saying

  • "This is exactly what an AI lab should be focused on: solving real, urgent problems for teams and businesses"Substack comment

  • "I tried a very simple one for presentations - I can see a lot of potential."Substack comment

  • "Unable to install"GitHub discussion

  • "How should developers record that a skill was tested in a specific environment?"GitHub discussion

  • "skill-security-review: audits other Agent Skills for safety"GitHub discussion

  • "PITH Inter-Agent Payload Compressor (Save 30-60% tokens)"GitHub discussion

GitHub70.9K

unslothai/unsloth

Local UI to run and train LLMs and diffusion models, including Qwen3.8, Kimi K3, MiniMax-H3, Gemma 4, DeepSeek-V4, FLUX and more.

youtube
Fast Fine Tuning with Unsloth

Market Signal

Why It Has Market Pull

Unsloth is a well-established, Y Combinator-backed open-source fine-tuning platform with real institutional credibility and a massive install base. It just extended its reach with a desktop app that lets non-technical users train and run models locally, generating significant social amplification including a reshare from a prominent tech figure.

  • Roughly 71,000 GitHub stars, 6,400 forks, and over 10 million monthly model downloads
  • Y Combinator alumnus, also backed by GitHub Accelerator, Microsoft's M12 fund, Redpoint Ventures, and Samsung NEXT
  • Nearly 800 open issues and 300+ open pull requests, evidence of a heavily used, actively maintained codebase
  • New desktop app was reshared by a widely-followed tech figure and amplified across X
  • Dual-licensed to sustain continued open development alongside the paid desktop UI

feedbacks

What People Are Saying

  • "It's the first desktop app that allows you to train and run models. You do everything with it, run + train diffusion image/video, serve models remotely, code execution, inference and more!"X reply

  • "Unsloth Desktop to run and train local models! Definitely gonna try it"X reply

  • "the installation took a long time, and it installed Python and other dependencies without warning or asking for permission"X reply

  • "users encountered installation failures on Windows, Mac, and AMD setups"aggregated X reports

  • "the startup taking on major labs in AI fine-tuning"trade press article

  • "Independent third-party coverage is still thin."evidence gap

GitHub8.9K

Lightricks/LTX-2

Official Python inference and LoRA trainer package for the LTX-2 audio–video generative model.

youtube
How To Use LTX-2 in ComfyUI | FREE AI Videos With Synced Audio

Market Signal

Why It Has Market Pull

LTX-2 is an open-source, DiT-based audio-video foundation model, one of the first open models to generate synchronized audio and video in a single pass at up to native 4K/50fps, released with full training code. Backed by a well-funded, established company, it drew a strongly positive community reception at launch and continues to gain stars and forks.

  • Roughly 8,900 GitHub stars and 1,400 forks
  • Built by a company that has raised roughly $335M at a reported $1.8B valuation from major growth investors
  • Its launch thread on a major AI-art community reportedly drew over 700 upvotes and 147 comments
  • Reported as developed with major cloud infrastructure backing
  • Full model weights, training code, and a LoRA trainer are open-sourced, a differentiated level of transparency versus closed competitors

feedbacks

What People Are Saying

  • "Independent third-party coverage is still thin."evidence gap

HF Spaces2.5K likes

Qwen-Image-Edit-2511-LoRAs-Fast

Demo of the Collection of Qwen Image Edit LoRAs Qwen-Image-Edit-2511-LoRAs-Fast is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 2495 likes on Hugging Face.

Market Signal

Why It Has Market Pull

This is the strongest performer among the Hugging Face demos reviewed: nearly 2,500 likes and 26 open community discussions, built by a prolific, well-regarded creator known for shipping reliable image-editing tools quickly. It bundles several distinct image-editing LoRA effects into one fast, actively maintained interface rather than a single-trick demo.

  • 2,495 likes, the highest of any Hugging Face Space reviewed here
  • 26 open community discussions point to sustained, ongoing engagement rather than a one-time spike
  • Built by a creator with a long track record of well-used Hugging Face image tooling
  • Combines multiple distinct LoRA effects into a single fast workflow instead of a narrow one-off demo
  • The owner is visibly responsive, patching UI issues and redirecting users to a faster companion Space within days of reports

feedbacks

What People Are Saying

  • "What happened to the interface on mobile devices? It's completely unusable."Hugging Face comment

  • "worst experience for regular users, drop down menu for style is not working on multiple browsers"Hugging Face comment

  • "The image results got cropped and stretched"Hugging Face comment

  • "keep getting Invalid repo token message even with a valid token?"Hugging Face comment

  • "The issue you're facing will be fixed soon. Until then, you can try this Space, which produces consistent results and even faster inference."Hugging Face comment (creator reply)

  • "Independent third-party coverage is still thin."evidence gap

Product Hunt537

Dograh

Closed voice platforms make you rent your own agents. Dograh is completely open source- nothing is gated. Visual flow builder, add your model key across 30+ integrations or use local models, telephony, human transfer, and advanced QA & monitoring - all free to self-host in one command. Also connect your claude code with MCP to build voice agents for a use case or call recordings.

youtube
Getting Started on Dograh

Market Signal

Why It Has Market Pull

Dograh is a fully open-source, self-hostable voice AI platform positioned as a direct alternative to closed voice-agent platforms, offering a visual flow builder, 30+ integrations, telephony, and native MCP support so it can be built and wired up by chatting with an AI coding agent. It was the top-ranked launch of its day and backs that up with a genuinely active open-source community.

  • 537 upvotes, finishing #1 for the day with 1,200 followers
  • GitHub repository has reached roughly 4,300 stars and is tracked as a trending repository
  • Also launched on Hacker News, drawing organic praise and specific technical feedback
  • Offers three ways to run, free self-hosted, managed cloud, and private VPC deployment, showing a real path to a sustainable business beyond the open-source core
  • Multiple independent YouTube walkthroughs position it directly against a closed competitor, indicating third-party interest beyond its own launch channels

feedbacks

What People Are Saying

  • "This is a great direction for voice AI!"Product Hunt comment

  • "CONGRATULATIONS first of all for getting this launched, I think I was the 100th one :)"HN comment

  • "Thank you for sharing your hard work with the world! I am grateful that you are helping us advance towards a real voice computer."HN comment

  • "It would be great to provide different voice personas. My agent reported some lag in getting responses during the call."HN comment

  • "Dograh Voice AI Tutorial - Open-Source Alternative"YouTube post

  • "I Tried Dograh as an Alternative: Voice Agent + n8n"YouTube post

Sources

GitHub

msitarzewski/agency-agents

145.1K

A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.

3b1b/manim

90.8K

Animation engine for explanatory math videos

infiniflow/ragflow

87.9K

RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs

kepano/obsidian-skills

45.6K

Agent skills for Obsidian. Teach your agent to use Obsidian CLI and open formats including Markdown, Bases, JSON Canvas.

altic-dev/FluidVoice

9.8K

Fastest and only macOS Dictation app with on-device STT and custom trained AI enhancement model. A local Wispr Flow alternative. ⭐ helps a ton :) Windows & iOS waitlist open. Linux soon.

Product Hunt

Grok Bot

493

Introducing Grok Bot: AI teammates you can give real work to. Bots sign into your tools, use them like you do, and come back with finished work. Message them like a teammate. They keep context, multitask, and learn your working style on their own cloud computer. Ask Bots to run outbound overnight, code expenses from Gmail, schedule a recruiting onsite, rebalance paid ads, or ship a bug pack. They only come back when something needs your approval.

Lettertrace

386

Lettertrace measures how often Claude, ChatGPT, and Gemini mention your company. But there's a catch: it's free, developer-first, bring-your-own-key, and open source.

Unsloth Desktop

229

Unsloth Desktop is an open-source app to run and train AI models locally. Run LLMs, image/video diffusion, and audio. Connect agents like Claude Code or Codex to your local GPU with one command, and fine-tune models with no-code workflows.

LaraCopilot

146

LaraCopilot is an agentic AI engineer that builds real, production-ready apps. Describe your idea in your own language and it builds the whole thing: UI, backend, database, auth, and APIs, wired together and ready to deploy. One-click deploy to the cloud or your own server, GitHub export, instant rollback. No lock-in. The code is yours.

CodeBurn

106

CodeBurn is a free, open-source tracker for AI coding costs. It reads the session files your tools already write - Claude Code, Cursor, Codex, Copilot, 40 in all, and shows every token and dollar by the task, model, project, and pull request that used it. The Optimize tab finds waste like cache bloat or retry tax, applies the fix, and tracks what it actually saved. Everything runs on your machine: no account, no uploads. MIT-licensed and genuinely free, used by 150k+ developers across the world.

Assembly Studio

257

Assembly Studio is the AI app builder made for professional service firms. Describe what you want: onboarding agents, analytics dashboards, community hubs, and get production-grade apps in minutes. 20+ templates to remix. Connects to any 3rd-party tool. Clone any overpriced app charging you per seat. Free forever plan. Stop paying for someone else's software. Build the firm only you can build.

YC Launch

Almanac: The agent with a company brain

12

Almanac works with its own browser and terminal, signed into the accounts you connect. Almanac · Summer 2026 · B2B Tags: Generative AI, AI, AI Assistant. Website: https://usealmanac.com

Risklytics: Insurance for the Frontier

8

Insurance for companies pushing the frontier of what is possible with AI, from robot fleets to agents in production. Risklytics · Summer 2026 · Fintech Tags: Artificial Intelligence, Hardware, Machine Learning, B2B, Insurance. Website: https://www.risklytics.ai/

Hacker News

Needle2: 14MB agentic LLM for phones, wearables, smart home and robots

527 pts

Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits betwee... (527 points, 181 comments).

Orca – Instant cross-platform in-app purchase orchestration

5 pts

Hi HN, Founder of Maxint here, we just launched Orca ( https://orca.maxint.com/ ) One major hurdle for us, following the deployment across iOS, Android, Linux, MacOS, Windows and Web, was monetization. There was no standardized store, schema or architecture to define, integrate and track subscriptions, consumables and non-consumables across platforms. We tried RevenueCat and Adapty, however both were limited to mobile and web, while we still had to manually set up all products and entitlements i... (5 points, 0 comments).

Codify – Terraform for Developer Environments

4 pts

Setting up developer environments is surprisingly manual. Most teams rely on a mix of docs, shell scripts, and tribal knowledge. We built Codify to make the process reproducible with a Terraform-like workflow. The project consists of an open-source CLI, a JSON5 based configuration language, a library of 50+ resources, a web/desktop editor, and an AI assistant. The workflow mirrors Terraform: plan + apply to make changes, and import + refresh to synchronize an existing configuration with the curr... (4 points, 5 comments).

HF Spaces

Wan2.2 14B Fast Preview [NEW]

947 likes

Wan2.2 14B Fast Preview [NEW] is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 947 likes on Hugging Face.

MiniMax H3

246 likes

Video generation with a synchronized soundtrack MiniMax-H3 — unquantized, split across two Spaces Joint video and soundtrack out of a single denoising pass, at bfloat16 with no quantization anywhere. This Space is the denoising half: the 61.73 GiB transformer and the two autoencoders. The 62.14 GiB Qwen3-VL conditioner runs in qwen3vl-conditioner, which this Space calls over the gradio API for every request. The weights are the public MiniMaxAI/MiniMax-H3 diffusers checkpoint. MiniMax-H3 is 195.9 GiB in bfloat16 and a ZeroGPU Space is evicted at 150 GB of storage. An unquantized single Space is therefore impossible, which is why quantized demos of it run NVFP4 or float8 weights. Cut the Mini...

CharacterSheet LoRA Demo

130 likes

Multi-view character sheet from one image (FLUX.2 LoRA) This Space demonstrates the CharacterSheet QuadView LoRA applied on top of FLUX.2 Klein 9B. Upload a clear, well-framed image of a character and the model produces a multi-view character sheet: a face close-up plus front, side, and back full-body views on a single 1536×1024 sheet. CharacterSheet LoRA Demo is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 130 likes on Hugging Face.

MiniMax H3 Turbo LoRA

86 likes

Video generation with a synchronized soundtrack MiniMax-H3 — unquantized, split across two Spaces Joint video and soundtrack out of a single denoising pass, at bfloat16 with no quantization anywhere. This Space is the denoising half: the 61.73 GiB transformer and the two autoencoders. The 62.14 GiB Qwen3-VL conditioner runs in qwen3vl-conditioner, which this Space calls over the gradio API for every request. The weights are the public MiniMaxAI/MiniMax-H3 diffusers checkpoint. MiniMax-H3 is 195.9 GiB in bfloat16 and a ZeroGPU Space is evicted at 150 GB of storage. An unquantized single Space is therefore impossible, which is why quantized demos of it run NVFP4 or float8 weights. Cut the Mini...

Wan2.2 14B Fast Preview

1.1K likes

generate a video from an image with a text prompt Wan2.2 14B Fast Preview is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 1122 likes on Hugging Face.

QIE-2511 Rapid-AIO LoRAs Fast (Experimental)

285 likes

Demo of the Collection of Qwen Image Edit LoRAs QIE-2511 Rapid-AIO LoRAs Fast (Experimental) is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 285 likes on Hugging Face.