Tools Bench.

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

Last Brew Time: Sep 10, 2026, 11:11 AM PT

Insight

Builders are treating the agent itself as a commodity and racing to own the layer that feeds it

01.PACKAGING AGENT SKILLS AS INSTALLABLE UNITS

02.SELF-HOSTING THE AI GATEWAY INSTEAD OF PAYING FOR ONE

03.BUILDING BESPOKE CODING-AGENT HARNESSES FROM SCRATCH

04.TRAINING PHYSICAL ROBOTS INSTEAD OF HAND-CODING THEM

Featured

GitHub35.1K

THU-MAIC/OpenMAIC

Open Multi-Agent Interactive Classroom — Get an immersive, multi-agent learning experience in just one click

github
OpenMAIC — open multi-agent interactive classroom generator

Market Signal

Why It Has Market Pull

OpenMAIC is Tsinghua University's open-source (AGPL-3.0) multi-agent classroom generator, turning a topic or PDF into an AI-taught interactive lesson. It has strong, verifiable open-source engagement, a credible academic backer, and real deployment evidence (700+ students, 100K+ learning records), making it one of the more substantively validated tools in this batch.

  • 35,198 real GitHub stars and 5,619 forks confirmed via GitHub API; 83 distinct contributors — a genuinely broad contributor base, not a single-author project
  • Actively developed with professional security hygiene: recent commits fix SSRF/cloud-metadata exposure and formalize a CVE disclosure process (Sept 2026)
  • Backed by Tsinghua University's MAIC lab; validated with 700+ real students and 100,000+ learning records, reporting 84.1% satisfaction per project materials, independently summarized by AllClaw
  • Covered by multiple independent outlets (AIToolly x2, AllClaw, aibase.com, Medium) rather than a single press release
  • v1.0.0 release added a 'Pro workbench' with durable course sessions and 20 built-in skills, showing continued product investment beyond initial launch

feedbacks

What People Are Saying

  • "transforms any topic description or PDF document into a complete, immersive multi-agent classroom in minutes"AIToolly

  • "Validated with over 700 real Tsinghua students and 100K+ learning records"AllClaw review

  • "84.1% overall satisfaction rate, significantly higher engagement and knowledge retention compared to traditional online lectures"AllClaw review

  • "a significant shift toward active, AI-powered education rather than passive video learning"aibase.com news

  • "fix(ssrf): keep cloud metadata endpoints blocked under ALLOW..."GitHub commit history

  • "docs(security): state the severity and CVE process for advisories"GitHub commit history

  • "OpenMAIC: Generate Interactive Virtual Classrooms with One Click"Medium (AI Engineering)

GitHub31.0K

vercel-labs/skills

The open agent skills tool - npx skills

youtube
How I Made Claude Code WAY Better With Vercel Agent Skills

Market Signal

Why It Has Market Pull

Skills is Vercel's open package manager for agent skills, letting a single `npx skills` command install reusable context across more than a dozen coding agents. Backing from a well-known company, a companion skills.sh directory, and rapid, sustained star growth make this one of the stronger market signals in this batch, though some users have questioned whether the leaderboard ranking favors Vercel's own skills.

  • 31,000+ real GitHub stars and 2,652 forks, closely matching the reported figure
  • Built and maintained by Vercel, announced via an official Vercel changelog post and backed by the companion skills.sh directory
  • 144 contributors and frequent releases, with v1.5.25 shipping September 8, 2026, plus a sibling vercel-labs/agent-skills library
  • Compatible with 15+ agent runtimes (Claude Code, Codex, Cursor, Windsurf, Amp, and more), positioning it as an ecosystem-wide package manager rather than a single-tool plugin

feedbacks

What People Are Saying

  • "What is this? How does it work? How are skills ranked? Seems a little bit fishy to me... despite there definitely being much more used skills in the overall AI coding ecosystem."HN comment

  • "The UI looks nice, otherwise."HN comment

  • "I just see an endless spinner."GitHub issue

  • "skills.sh still renders stale SKILL.md after source update + fresh reinstall"GitHub issue

  • "Official support for private skills with documentation"GitHub feature request (30 comments)

  • "skills are becoming a standard... a much bigger deal long-term"HN comment

HF Spaces494 likes

Microduck Sandbox

Real trained RL policies for the Microduck robot, running fully in the browser: MuJoCo compiled to WebAssembly steps the physics, onnxruntime-web runs the policy network at 50 Hz. No server, no backend. Two locomotion variants of the same robot are included: legs (walking, the default) and rollers (the wheeled skating variant). Press M (or hold D-pad up ~1 s on a gamepad) to switch; the roller model, meshes and policies are lazy-loaded on the first switch. | Mode | Checkpoint | What it does | |--------|-----------|--------------| | Run (legs) | BESTalphawalking.onnx | Velocity-tracking locomotion (arrows / WASD to steer) | | Sit | BESTalphasitstand.onnx | Sits down on its hull, stands back u...

github
Microduck — tiny biped robot driven by reinforcement-learning policies

Market Signal

Why It Has Market Pull

This is the official in-browser simulator for Pollen Robotics and Hugging Face's Microduck, a real $399 open-source reinforcement-learning biped robot, running actual trained ONNX policies via MuJoCo compiled to WebAssembly. It has substantial press coverage, an active open-source training pipeline, and a growing community-tooling ecosystem, making it a strong, differentiated, brand-safe pick for a professional AI/robotics builder audience.

  • Official pollen-robotics organization Space, 494 HF likes
  • Backs a real shipping product: a $399 open-source 25cm biped robot trained via RL
  • Runs real trained policies in-browser at 50Hz via MuJoCo WASM + onnxruntime-web, with gamepad support
  • Covered by MarkTechPost, Zeli, and other outlets within weeks of launch (late Aug 2026)
  • Spawned an independent community-curated GitHub list (joeynyc/awesome-microduck) and a from-scratch React Three Fiber community rebuild (MicroDuckModels)

feedbacks

What People Are Saying

  • "Hugging Face Unveils Microduck: A $399 Open-Source 25 cm Biped You Train with Reinforcement Learning"MarkTechPost

  • "MuJoCo compiled to WebAssembly plus onnxruntime-web running the real policies at 50 Hz, with gamepad support"project documentation

  • "A curated list of software, simulators, policies, agent tools and coverage for the Pollen Robotics / Hugging Face Microduck robot"GitHub (awesome-microduck)

  • "roughly one to two hours on a CUDA GPU for a usable gait at 4096 parallel environments"microduck_rl training docs

  • "Microduck: A 25 cm open-source biped robot you train with reinforcement learning"Zeli coverage

Product Hunt147

AlphaGenome Atlas

AlphaGenome Atlas is Google DeepMind's AI-powered map of how genetic mutations may affect human biology. Built by precomputing AlphaGenome predictions for all 9 billion possible single-letter DNA changes, the 1-petabyte dataset lets researchers explore and prioritize variants across both coding and non-coding regions. It's free to explore through a visual web interface, with API and Antigravity access for deeper research.

youtube
AlphaGenome: AI for better understanding the genome

Market Signal

Why It Has Market Pull

AlphaGenome Atlas is Google DeepMind's searchable, precomputed map of predicted molecular effects for all 9 billion possible single-letter human DNA changes, released as a free web portal, API, and Google Antigravity skill. It drew coverage from Nature, Scientific American, and Fortune and was described by commenters as an 'AlphaFold moment' for genomics, reflecting genuine scientific and market weight behind the launch.

  • 147 Product Hunt votes plus front-page Hacker News discussion across two separate threads
  • Backed by a 1-petabyte dataset, described as 30x larger than the AlphaFold database
  • Covered by Nature, Scientific American, Fortune, and MarkTechPost as a major research release
  • Open, code-accessible via github.com/google-deepmind/alphagenome and alphagenome_research repos plus a public API
  • Comments compare it directly to AlphaFold's field-changing impact, a strong credibility signal from a technical audience

feedbacks

What People Are Saying

  • "This really does feel like an AlphaFold moment."Product Hunt comment

  • "Now I'm just curious what unexpected stuff people find with this."Product Hunt comment

  • "9 billion DNA variants mapped and free to explore, no coding needed."Product Hunt comment

  • "Models of this type either release a precomputed database, or somebody else releases one for it, or it gets ignored... the scale, the coverage of the non-coding 98%, and the attributions"Hacker News comment

  • "New Google DeepMind atlas could transform our understanding of genetic diseases"Scientific American headline

  • "DeepMind's new genome 'atlas' charts effects of all nine billion possible mutations"Nature coverage

Sources

GitHub

obra/superpowers

284.6K

An agentic skills framework & software development methodology that works.

diegosouzapw/OmniRoute

64.0K

Never stop coding. Free MIT AI gateway: one endpoint, 352 providers (150+ free), 1200+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 550+ contributors

ayghri/i-have-adhd

37.6K

A skill to stop your coding agent from burying the answer. ADHD-friendly output.

cathrynlavery/diagram-design

37.5K

38 editorial diagram types for Claude Code, Codex, and Pi. Self-contained HTML + SVG. No shadows. No Mermaid slop.

AlexsJones/llmfit

35.6K

Hundreds of models & providers. One command to find what runs on your hardware.

freestylefly/awesome-gpt-image-2

30.8K

Prompt as Code | GPT Image 2 / 2.5 提示词与案例库，530+ 个案例、20+ 套工业级模板与可复用 Skills，新增 2.5 同提示词对比专区，附完整提示词与生成记录，持续更新。

Product Hunt

Mastra Factory

474

Mastra Factory is an open source, agent-powered software delivery environment. It combines persistent coding agents, repository workspaces, issue intake, planning, implementation, and pull request review in a web application you control.

Harden

397

Harden AIF is a free, local security tool for AI coding agents. Its post-trained model checks tool calls before they run, using your request and session context. It beat frontier models on key agent-security benchmarks, while keeping your repo and tool output on your machine.

ChatGPT Images 2.5

305

ChatGPT Images 2.5 is OpenAI's next-generation image model built for sharper details, faster generation, and more precise editing. It helps you turn sketches, prompts, and reference photos into polished visuals with better consistency, natural lighting, richer textures, and stronger control across every edit.

Noodle Seed

249

Noodle Seed helps software teams make their products ready for AI agents. Build workflows in TypeScript, expose them through a secure branded assistant inside your product, and make the same capabilities available to external agents. Instead of stitching together MCP SDKs and hosting infrastructure, Noodle Seed provides the governed runtime for identity, permissions, secrets, audit, and operations.

Muse by Meta

224

Try Muse, your personal AI agent that gets things done. Give Muse a goal or an everyday task and it handles the rest, from finances and health to shopping and the people you care about.

DuckFightClub

141

Imagine WWE SmackDown, but with cute, AI-trained robots. Eight teams spend days training reinforcement-learning policies for Pollen’s open-source MicroDuck, then we throw them in the simulator to fight for the Golden Beak Belt. The whole competition will be livestreamed. The catch: you’re not building a robot, you’re training its brain. Want in? Register a team, train your policy, or host a showdown in your city. When the real MicroDucks ship, we go IRL.

YC Launch

Okibi - Agent-ready CLIs

16

Turn your product into a CLI, and agents into customers Okibi · Summer 2025 · B2B Tags: Developer Tools, SaaS, AI. Website: https://okibi.ai

Hacker News

ShevtoneAudio Orchestrator – Turning MIDI into Full Orchestration

6 pts

Hi HN, I'm Bor Shev, a composer and developer. Over the past two years I've been developing ShevtoneAudio Orchestrator. The idea is simple: instead of generating a finished piece of music and replacing the composer, Orchestrator takes the composer's own MIDI and develops it into a full orchestration. It analyzes the musical material — harmony, melody, rhythm, dynamics, structure and orchestral density — and creates an arrangement across strings, brass, percussion and other sections. The importan... (6 points, 4 comments).

Open-source simulation testing infra for voice agents

5 pts

Hey HN, we’re Nischal & Naman. We’re brothers, and together we’re building an open-source platform for simulation based testing of voice agents (try it out in 5 mins - https://docs.egma.ai/docs/get-started/quickstart , 2 min demo video - https://youtu.be/wgDWEe5UAUY ) Platforms that help you do simulation testing already exist. But they all charge a heavy premium on top of inference costs. We believe if the industry truly wants to scale simulation testing of voice agents, we need to stop chargin... (5 points, 1 comments).

Syq – copy files between machines fast (better than rsync)

17 pts

I got frustrated with the slowness of rsync and made an alternative that works faster by using multiple parallel connections, direct encrypted TCP when available, and other optimizations. I also added cool features like the ability to maintain a persistent ssh connection to the server for fast one-offs, the ability to download to your laptop while working in an ssh shell on a server, and the ability to do direct remote-remote transfers without forwarding your ssh agent (by using restricted ssh k... (17 points, 19 comments).

Rubato – Retro-Mac desk device mirrors AI coding state-ESP8266

7 pts

Hi everyone! I have built an AI health device (ESP8266 + 240*240 screen). It can show AI Agents' status in real time with breathing bubble. And remind user dringking water, toilet, streth, etc. support: DeepSeek Harness, opencode, OpenClaw, Claude Code, Cursor. Hardware design, firmware and plugins are all opensource. urls are here: https://github.com/lovaxi/Rubato_Device https://github.com/lovaxi/Rubato_Plugins Type-c power supply, 2.4G wifi. Online sell on Tindie: https://www.tindie.com/produc... (7 points, 2 comments).

HF Spaces

MiniMax H3 Turbo LoRA

409 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...

H3 Acceleration Arena

88 likes

Blind A/B ranking of MiniMax-H3 acceleration variants Human-judged ranking of ~26 MiniMax-H3 acceleration variants over a 200-prompt corpus, from blind pairwise votes on pre-generated clips, with confidence intervals, cost and slice breakdowns. The design and its reasoning are in arena/DESIGN.md; the app's own notes are in arena/README.md. This Space is private and must stay private until deliberately flipped. It streams ~3,700 clips out of the private dataset multimodalart/h3-pre-gen-arena. See Going public below. hfoauth: true above creates the OAuth app and injects OAUTHCLIENTID, OAUTHCLIENTSECRET, OAUTHSCOPES and OPENIDPROVIDERURL. arena/space_auth.py implements the flow by hand (this is...

Fruit Fly Simulation

51 likes

Simulate a fruit fly in your browser using WebGPU kernels A standalone fruit fly connectome demo: paint neurons, stimulate the network, and watch an articulated Three.js fly respond. Vanilla JavaScript, Vite, Three.js, and @huggingface/kernels. Simulated neural activity drives crafted walking, turning, and flight animations. The movements are illustrative, not validated predictions of fly behavior. Open the address printed by Vite, then click Download & start. All neural weights, fly meshes, fonts, and kernel templates are included in public/; setup loads them into the browser and caches verified weight chunks. No API key or external model service is required. Deploy the generated dist/ dire...

Viggle-Animate

44 likes

Put the person from a still into a driving video, in 4 steps Upload a driving video and a character still. The clip's first frame is repainted with gpt-image-2 so it shows your character in that exact pose and framing, and that repainted frame is the only conditioning the model gets besides the clip itself — no pose estimation, no segmentation, no masks. Takes 4–10 s of driving video and renders it at 4 sampling steps; anything longer is cut to 10.1 s. The model itself only renders frame counts of the form 17k+5 and never fewer than 124 (5.17 s), so shorter clips are held out to 124 frames with a frozen last frame and the render is trimmed back — what comes out is exactly as long as what wen...

Wan2.2 14B Fast Preview

2.0K 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 1956 likes on Hugging Face.

FLUX.2 Klein multi-LoRA

543 likes

Use multiple FLUX.2-Klein LoRAs Note: This space is experimental and may log image-uploads during certain periods for performance monitoring. Always comply with HF and model Terms of Service. Any stored images are automatically deleted after 7 days. FLUX.2 Klein multi-LoRA is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 543 likes on Hugging Face.

AI Tools — September 10, 2026 Edition | Agentic Brew