AI tools, agents & startups on the rise

Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.

What it does

Strata runs Qwen3.8-Flash-Next on consumer hardware with a one-click install for Windows and Linux, a local OpenAI and Anthropic API, and optional image input.

What's unique

One-click local inference of Qwen3.8-Flash-Next, with a localhost API and optional images.

Why it helps

People without datacenter GPUs can run that Qwen model locally and call it through a familiar API.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: 22 days old · 24k ⭐ on GitHub · 932 points on Hacker News.

An agent skill for creating detailed, interactive isometric illustrations of real-world machines and complex systems. Generates polished SVGs with optional WebGL shaders and 3D effects.

What it does

Anatomy is an agent skill that creates interactive isometric illustrations of machines and complex systems, as polished SVGs with optional WebGL shaders and 3D effects.

What's unique

It is an open-source agent skill aimed at isometric SVG illustrations rather than a general image generator.

Why it helps

People explaining a machine can ask an agent for a detailed isometric figure they can inspect.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: Agent pick · 8 days old · 14k ⭐ on GitHub.

Browser and phone client for herdr: chat and live terminal for every agent pane, remote PCs over SSH, web push alerts

What it does

herdr-web-ui is a browser and phone client for herdr, with chat and a live terminal for every agent pane.

What's unique

It reaches remote PCs over SSH and sends web push alerts.

Why it helps

You can watch coding agents from a phone and get a push when they need you.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: 28 days old · 1.1k ⭐ on GitHub · +386 ⭐ this week.

Code graph and memory for your coding agents, in one Rust program. For Claude Code, Codex, Cursor and any MCP client.

What it does

A single Rust program that builds a code graph and project memory for coding agents.

What's unique

It works with Claude Code, Codex, Cursor, and any MCP client.

Why it helps

Coding agents can use one local graph and memory store for cited context from a repository.

Pricing

Open source (Apache-2.0) — free to self-host

Why it’s on the radar: Agent pick · 2 days old · 184 ⭐ on GitHub · 40 points on Hacker News.

An open-source AI accelerator, developed by AI: RTL, ISA, simulator, compiler and profiler in one repo. Runs Qwen3, LFM2.5 and Qwen3.5 on a Kintex-7 PCIe card.

What it does

openTPU is an open-source AI accelerator from AI, with RTL, an ISA, a simulator, a compiler, and a profiler in one repo.

What's unique

It runs Qwen3, LFM2.5, and Qwen3.5 on a Kintex-7 PCIe card.

Why it helps

Hardware researchers can study or extend a full accelerator stack that already runs those models.

Pricing

Open source (Apache-2.0) — free to self-host

Why it’s on the radar: 22 days old · 749 ⭐ on GitHub · 343 points on Hacker News.

Reverse engineer anything with agents, from app behavior down to native binaries.

What it does

rea reverse engineers anything with agents, from app behavior down to native binaries.

What's unique

Agents can go from how an app behaves all the way down to native binaries.

Why it helps

Developers can ask an agent to explain an app's behavior and its native binaries.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: Older project (since 2026) gaining momentum · 287k ⭐ on GitHub · 63 points on Hacker News.

AI Agent Builder and Runtime by Docker Engineering

What it does

docker-agent is an AI agent builder and runtime from Docker Engineering.

What's unique

The agent builder and runtime come from the Docker engineering team.

Why it helps

Teams already on Docker can build and run agents in a runtime from Docker Engineering.

Pricing

Open source (Apache-2.0) — free to self-host

Why it’s on the radar: Older project (since 2025) gaining momentum · 4.5k ⭐ on GitHub · 300 points on Hacker News.

Search your computer by what's in your files: photos, PDFs, documents, recordings, videos and code. On-device on Mac, Windows and Linux, built on Google DeepMind's EmbeddingGemma 2.

What it does

DigUp searches a computer by what is inside photos, PDFs, documents, recordings, videos, and code, on device for Mac, Windows, and Linux.

What's unique

It is built on Google DeepMind's EmbeddingGemma 2 and stays on the machine.

Why it helps

People who want file search without uploading content can look across local files on Mac, Windows, or Linux.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: 7 days old · Featured in Ben's Bites · 4.3k ⭐ on GitHub.

Let your AI agents paint big arrows, boxes and text on your Mac screen. One CLI, click-through, gone by itself. Skill for Claude Code and Codex. MIT.

What it does

big-arrow-on-the-screen lets AI agents paint big arrows, boxes, and text on a Mac screen from one CLI. The overlay is click-through and goes away by itself.

What's unique

It is a Claude Code and Codex skill, MIT licensed, that draws on the screen without blocking clicks.

Why it helps

People steering agents on a Mac can point at the UI without the overlay getting in the way.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: 8 days old · 764 ⭐ on GitHub · 247 points on Hacker News.

Your personal AI agent in your own Cloudflare account. Chat, memory, tasks, notes and scheduled reminders, deployed with one command: npx create-talorys@latest. Free-tier friendly, single-user, no telemetry.

What it does

Talorys is a personal AI agent that deploys into your own Cloudflare account with one command.

What's unique

It includes chat, memory, tasks, notes, and scheduled reminders, and is free-tier friendly, single-user, and without telemetry.

Why it helps

You get a self-hosted assistant in your Cloudflare account without a separate telemetry service.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: 6 days old · 1.2k ⭐ on GitHub · 188 points on Hacker News.

OpenShell is the safe, private runtime for autonomous AI agents.

What it does

OpenShell is a runtime for autonomous AI agents that is meant to stay safe and private.

What's unique

It focuses on a safe, private place for agents to run on their own.

Why it helps

Developers who want autonomous agents can run them in a runtime built for safety and privacy.

Pricing

Open source (Apache-2.0) — free to self-host

Why it’s on the radar: Older project (since 2026) gaining momentum · Featured in Ben's Bites · 16k ⭐ on GitHub.

Jeeves – Reasoning improves Jev-like decision models

What it does

Jeeves applies reasoning to improve Jev-like decision models.

What's unique

It targets decision models in the Jev style rather than general chat.

Why it helps

People building Jev-like decision models can use Jeeves to improve their reasoning.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: 17 days old · 417 ⭐ on GitHub · 242 points on Hacker News.

Open source inference engine for agents that optimizes itself for your exact hardware. Compiles and tunes its kernels on your device, so open models run up to 2x faster than llama.cpp. Works on Apple Silicon, NVIDIA, AMD, or just a CPU.

What it does

magnitude is an open-source inference engine for agents that compiles and tunes kernels on your own hardware.

What's unique

It claims open models run up to 2x faster than llama.cpp on Apple Silicon, NVIDIA, AMD, or CPU.

Why it helps

Agent developers can get hardware-tuned inference without hand-writing kernels for each machine.

Pricing

Open source (Apache-2.0) — free to self-host

Why it’s on the radar: Older project (since 2026) gaining momentum · 6.6k ⭐ on GitHub · 194 points on Hacker News.

Open source SDK to build Muse gadgets

What it does

muse-gadget-sdk is an open-source SDK for building Muse gadgets.

What's unique

It is specifically an SDK for Muse gadgets, and it is open source.

Why it helps

Developers can build their own Muse gadgets with the open SDK.

Pricing

Open source (Apache-2.0) — free to self-host

Why it’s on the radar: 13 days old · Featured in Ben's Bites · 2.2k ⭐ on GitHub.

在 dsh 里装上这个插件即可，无需登录、注册或填 API Key，就能使用包括 DeepSeek V4.1 Flash、Kimi K3 在内的前沿模型——完全免费，不限量。 All you do is install this plugin in dsh: no login, no sign-up, no API key — the frontier models are just there, DeepSeek V4.1 Flash and Kimi K3 among them. Completely free, with no usage cap.

What it does

This dsh plugin serves frontier models, including DeepSeek V4.1 Flash and Kimi K3, with no login, sign-up, or API key. It is completely free with no usage cap.

What's unique

Install the plugin in dsh and the models are available with no account and no usage cap.

Why it helps

dsh users can call DeepSeek V4.1 Flash and Kimi K3 without an API key or a usage limit.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: 22 days old · 19k ⭐ on GitHub.

Every agent's model. One place. Codex on DeepSeek, Claude Code on Kimi, from the menu bar.

What it does

magpie puts every agent's model in one menu-bar place, including Codex on DeepSeek and Claude Code on Kimi.

What's unique

You can point Codex at DeepSeek and Claude Code at Kimi from the menu bar.

Why it helps

Developers can swap which model an existing coding agent uses without leaving the menu bar.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: 23 days old · 14k ⭐ on GitHub.

Newsletters where you pay for the emails you send, not for dormant subscribers, at a fraction of the price of beehiiv and the like. Open source: self-host it, or use mailcheap.ai.

What it does

An open-source newsletter for the agentic era that you can self-host or use as a hosted service.

What's unique

It is described as the affordable newsletter for the agentic era, and the project is open source.

Why it helps

Publishers who want an agent-era newsletter can self-host it or use the hosted option.

Pricing

Open source (AGPL-3.0) — free to self-host

Why it’s on the radar: 6 days old · Featured in Ben's Bites · 1.1k ⭐ on GitHub.

The chat decision assistant: before you reply, Jev reads the chat, judges intent and risk, and drafts replies you fill in with one tap. You press send. Android · Windows · macOS · iOS | 聊天辅助决策工具：先看懂对方再回复，发不发由你。

What it does

jev-chat-jarvis reads a chat, judges intent and risk, and drafts a reply you fill in with one tap. You still press send. It runs on Android, Windows, macOS, and iOS.

What's unique

Jev drafts the reply, but sending stays a human action across phone and desktop.

Why it helps

People who want a second look at a chat can get a drafted reply without auto-sending it.

Pricing

Open source (MIT) — free to self-host

Why it’s on the radar: 25 days old · 7.8k ⭐ on GitHub.