Tools Bench.

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

Last Brew Time: Jun 21, 2026, 7:25 AM PT

Insight

The bottleneck builders are attacking isn't intelligence, it's what a model can remember and afford

Featured

GitHub72.3K

An open-source long-horizon SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of tasks that could take minutes to hours.

Market Signal

Why It Has Market Pull

DeerFlow is ByteDance's open-source SuperAgent harness that gives each agent its own sandboxed computer to research, code, and create across multi-hour tasks. Backed by ByteDance and propelled to the top of GitHub Trending on launch, it has become one of the most-watched agent frameworks of the year.

  • 72,338 GitHub stars and 9,811 forks since the May 2025 launch
  • Roughly 280 contributors with same-day commit activity
  • Reached the top of GitHub Trending within 24 hours of the 2.0 open-sourcing
  • Model-agnostic: runs OpenAI, Anthropic, Google, DeepSeek and ByteDance Doubao via YAML config
  • Each agent runs in an isolated Docker sandbox with filesystem, shell, and package install

feedbacks

What People Are Saying

  • "The open-source AI agent that actually works."YouTube video title

  • "FULLY free local deep research agent - it can do almost anything."YouTube video title

  • "Responses sometimes cut off mid-stream and fail to render."Hands-on review

  • "I couldn't find a scenario where it was the only viable option versus Claude Code."Hands-on review

  • "Schema migrations aren't wired up: create_all can't add columns to existing DBs."GitHub issue

  • "Docs introduction page returns a 404."GitHub issue

GitHub18.5K

Cognee is the open-source AI memory platform for agents. Give your AI agents persistent long-term memory across sessions with a self-hosted knowledge graph engine.

Market Signal

Why It Has Market Pull

Cognee is the open-source AI memory platform that turns documents, chats, and media into a self-improving knowledge graph agents can recall and reason over. Berlin-based and venture-backed, it pairs an actively released open-source engine with a managed cloud for teams.

  • Raised a $7.5M seed led by Pebblebed, with 42CAP and Vermilion Ventures
  • Angel backing from Google DeepMind, n8n, and Snowplow operators
  • 18,482 GitHub stars and 1,962 forks, with roughly 185 contributors
  • v1.2.0 released on a steady cadence (v1.1.3 three days earlier)
  • Building a Rust engine for on-device and edge memory alongside the cloud platform

feedbacks

What People Are Saying

  • "Superior AI memory and knowledge for AI agents - greatly beats a stock chatbot."YouTube video title

  • "Pick Cognee when entities and relationships are first-class concerns."Vectorize comparison article

  • "A memory layer you bolt onto your own agent."Comparison article

  • "RAG can demonstrate ideas but cannot support production workloads alone."Memgraph blog

  • "API error handlers return raw exception strings to clients."GitHub issue

  • "ALLOW_CYPHER_QUERY defaults to true; the user query is passed straight to the graph engine."GitHub issue

HF Spaces16.6K likes

Generate any application by Vibe Coding it DeepSite is a Vibe Coding Platform designed to make coding smarter and more efficient. Tailored for developers, data scientists, and AI engineers, it integrates generative AI into your coding projects to enhance creativity and productivity. DeepSite v4 is a Hugging Face Space tagged with docker, region:us. It has 16617 likes on Hugging Face.

Market Signal

Why It Has Market Pull

A breakout, top-tier vibe-coding tool that turns a single prompt into a working web app, riding the open-model wave with endorsements from Hugging Face leadership.

  • Roughly 16,600 likes place it among the most-liked tools on the entire Hugging Face platform
  • Hugging Face co-founder publicly called it 'totally insane for vibe-coders', amplified across X and LinkedIn
  • Now on its v4 release, with v2->v4 shipped on a steady cadence rather than as a one-off demo
  • Runs as a live, hosted demo on GPU hardware powered by open DeepSeek models
  • Covered by independent reviewers as the fastest path from prompt to a working single-page app

feedbacks

What People Are Saying

  • "The new DeepSite space is really insane for vibe-coders."HF community

  • "DeepSeek vibe coding with DeepSite is going viral with awesome projects!"HF community

  • "Unlike ChatGPT, you don't have to feed it a complex prompt to make one page look decent."Blog review

  • "Build a full website with one prompt - live demo."YouTube video title

  • "Great for simple sites; if you want more than one page, it can't do that yet."Blog review

  • "These more recent versions feel less polished than the original."HF community

  • "DeepSite not working for me right now."HF community

Product Hunt180

Meet Mellum, a family of fast language models, including a next-generation model for ultra-low-latency and high-performance inference.

Market Signal

Why It Has Market Pull

Mellum is JetBrains' open-weight family of fast code models, purpose-built for low-latency completion. Backing from an established IDE vendor plus genuine Hugging Face adoption makes this the most usage-proven launch of its cohort.

  • Made by JetBrains (IntelliJ, PyCharm, Kotlin) and open-sourced under Apache 2.0 across all variants
  • Real adoption on Hugging Face: the Mellum2 Thinking model draws ~25,600 downloads/month with 300 likes
  • Original 4B base has 451 likes and 12 community quantized variants
  • Reached #4 Product of the Day on Product Hunt with ~181 upvotes
  • Ships multiple weights (4B base plus Python/Kotlin fine-tunes and a 12B MoE) with official GGUF quants for local use

feedbacks

What People Are Saying

  • "Latency-first models are underrated; a focused in-IDE model is a smart bet."Product Hunt comment

  • "Shipping a smaller coding model as open weights is the interesting part - FIM quality is what matters."Product Hunt comment

  • "We need more of this to remove the dependency on cloud-based models."Product Hunt comment

  • "How does it compare with the latest Qwen and Gemma coding models?"Product Hunt comment

  • "How do I wire it into PyCharm autocompletion, or VS Code?"Hugging Face

  • "Best suited for engineers and researchers who want to fine-tune domain-specific code models."InfoWorld article

Sources

GitHub

Skills for Real Engineers. Straight from my .claude directory.

Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface

Penpot: The open-source design tool for design and code collaboration

Extracted system prompts from Anthropic - Claude Fable 5, Opus 4.8, Claude Code, Claude Design. OpenAI - ChatGPT 5.5 Thinking, GPT 5.5 Instant, Codex. Google - Gemini 3.5 Flash, 3.1 Pro, Antigravity. xAI - Grok, Cursor, Copilot, VS Code, Perplexity, and more. Updated regularly.

754 structured cybersecurity skills for AI agents · Mapped to 5 frameworks: MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, D3FEND & NIST AI RMF · agentskills.io standard · Works with Claude Code, GitHub Copilot, Codex CLI, Cursor, Gemini CLI & 20+ platforms · 26 security domains · Apache 2.0

Product Hunt

Meet the AI team for your team. WorkClaws are collaborative, proactive AI coworkers who work in Slack and Microsoft Teams just like every other colleague. They're fully customizable to get work done your way by learning skills and routines. Unlike most AI products that pair one person with one assistant, WorkClaws can collaborate 24/7 with your whole team. Each Claw has a job title, a manager in your org chart, and a cloud-hosted ClawOS computer with the ability to access more than 3,000 apps.

Slackbot’s new MCP Client ends fragmented AI work by connecting 20+ apps (Atlassian, Linear, Canva, Zoom) to one conversational interface. Ask Slackbot in plain language to act across tools—sign docs, update tickets, view dashboards—then share results in team channels for true multiplayer collaboration.

Most AI agent workflows lose useful context between sessions, tools, and chats. The usual fixes are either too manual, like copying notes into docs, or too heavy, like setting up a database, vector store, or custom RAG stack. pumaDB gives agents a simple shared place to save and reuse notes, facts, preferences, project context, transcripts, task state, and other useful memory. No database setup, vector DB, or infrastructure to manage

Basedash now has groups and access controls. Bundle users into groups — internal teams, external clients, leadership — and give each one access to exactly what it needs: data sources, MCP servers, dashboards, chats, and automations. Set a group's AI context so the assistant answers differently per audience, and row-level security applies to every question. The right people see the right data.

Reframe is an open-source browser based on Electron for macOS that brings back the look & feel of Safari 1.0, Netscape 4.8, Firefox 1.0 and Internet Explorer 5.0, with a built-in Wayback Mode.

The weights are the billions of numbers forming an AI's brain. Type a name and see how strongly the leading AI models recognize it. Are you in the weights?

Hacker News

I'm the developer of Trace, a non-intrusive, shortcut-driven Mac app that records and transcribes your meetings on-device. I know, another meeting transcription app. Please bear with me though, I'm confident that this is at least a little novel. I primarily built Trace for myself. I'd been using MacWhisper, but there was enough fiddling before each call that I'd forget to start it and walk out of an hour-long meeting with nothing written down. So the things I cared about most were that it's quic... (205 points, 83 comments).

Anthropic and OpenAI's publicly available models are explicitly guard-railed so that they refuse offensive tasks. And their cyber-focussed models are gated for enterprises. This leaves SMEs and mid market open to major vulnerabilities. AI can be used as both an adversarial and defensive tool in the world of cyber. A worst case outcome is if only the adversaries have access. Meanwhile, most existing AI cyber tools are just wrappers. The problem is that they still have all the guardrails on from t... (84 points, 38 comments).

Hi HN, I'm Kaushik, and I built Rocketgraph. I believe that while other spaces have caught up to the AI wave, the observability space is still lagging behind, using the same tools and dashboards that we use to analyse logs from human-written code. But now the code is written and debugged by AI, so we need to rethink how we do observability where the observer itself is an AI. The problem that I run into is when an alert fires, I have to manually check the Grafana dashboards and write LogQL querie... (13 points, 3 comments).

Hey everyone. My cofounder and I are formally open sourcing (MIT) persona.js. It's a VanillaJS library that helps anyone build agentic experiences on the web, without a framework dependency, and full WebMCP support So, why'd we do this? 1) We're super fans of the web and the browser can do a ton today 2) We've seen AI builds be way overly complex because the FE requires a large project within an existing app OR the site wasn't using a framework to begin with If you've been a part of huge apps wi... (11 points, 14 comments).

I'm Evan and I made Ito.ai It's code review that actually runs your code. The result is that it finds more bugs with a smaller false positive rate. You get screenshots, videos, and run logs. I built Ito because manual verification sucks and I thought I could automate it. I've been a professional developer for 15 years and now more than ever, I'm just spending time making sure things work. I looked into all of the other agentic testing tools that click around a url you give them and frankly, they... (10 points, 8 comments).

Hi HN, I'm the author of git-lrc, would appreciate some feedback from the community Last year my team started using AI coding tools more heavily, and we found ourselves generating tons of code, but spending less time looking at the stuff that's been generated. We felt like we were losing a bit of grip/understanding of what we were building. Regressions occasionally slipped through. Sometimes changes made it all the way to production only to be reverted later. We tried several AI code review tool... (9 points, 0 comments).

HF Spaces

11.1K likes

Create your own AI comic with a single prompt Last release: AI Comic Factory 1.2 The AI Comic Factory has an official website: aicomicfactory.app For more information about my other projects please check linktr.ee/FLNGR. If you like the AI Comic Factory, let me know! I am always creating new spaces and exploring new ideas for demos, meaning I don't have much time to take care of all of them (I wish I could clone myself or ask robots to do it). If you appreciate the AI Comic Factory and would like to leave a tip, that would be very kind 🫶 First, I would like to highlight that everything is open-source (see here, here, here, here). However the project isn't a monolithic Space that can be dupli...

Kolors Virtual Try-On is a Hugging Face Space tagged with gradio, region:us. It has 10111 likes on Hugging Face.

5.1K likes

Wan2.2 Animate is a Hugging Face Space tagged with gradio, region:us. It has 5113 likes on Hugging Face.

5.1K likes

AudioCraft is a PyTorch library for deep learning research on audio generation. AudioCraft contains inference and training code for two state-of-the-art AI generative models producing high-quality audio: AudioGen and MusicGen. Installation AudioCraft requires Python 3.9, PyTorch 2.0.0. To install AudioCraft, you can run the following: We also recommend having ffmpeg installed, either through your system or Anaconda: At the moment, AudioCraft contains the training code and inference code for: MusicGen: A state-of-the-art controllable text-to-music model. AudioGen: A state-of-the-art text-to-sound model. EnCodec: A state-of-the-art high fidelity neural audio codec. Multi Band Diffusion: An EnC...

The ultimate guide to training LLM on large GPU Clusters Instruction to install and run locally Loading HTML fragments: There are two way to load HTML fragments: Compile them into html during build time Fetch them and insert them during run-time When to use what Use compile time fragments only on parts which you want to ensure are seen by every user right after page load (e.g logo) Use run-time fragments for everything else so that the final HTML is of reasonable size (<1MB idealy) How to add a new fragment Add it to the src/fragments folder (e.g. src/fragments/banner.html) For run-time fragments, add {{{fragment-name}}} to appropriate place in src/index.html (e.g. {{{fragment-banner}}}) For...

3.8K likes

Apply the motion of a video on a portrait Live Portrait is a Hugging Face Space tagged with gradio, Multimodal, Motion control, Image-to-Video, Video-to-Video. It has 3752 likes on Hugging Face.

AI Tools — June 21, 2026 Edition | Agentic Brew