Top Dev Tool Projects : Playwright, CapRover, Santa, Octane, Webiny & Project AIRI Transcript, AI Summary & Key Points

00:00 Every week developers release powerful open-source dev tools, and this weekly dev tool project update video brings them together in one place. Top trending open-source and best dev tool projects this week cover coding agent skills, self-hosted deployment platforms, security scanners, web testing frameworks, and charting libraries. You'll discover useful and trending developer tools you can start using right away, explained fast and to the point.

00:23 Without wasting time, let's get started. >> Before we jump into today's project updates, here's a quick announcement for everyone. We've launched a brand new YouTube channel called AI Agent Studio, dedicated entirely to AI agent projects, tutorials, and tools. So, if you're interested in staying up to date with the latest AI agent open-source projects, learning how to build your own agents, or exploring cutting-edge agent frameworks, make sure to check it out.

00:51 Subscribe now to get weekly videos, in-depth guides, and real-time project breakdowns. The link is right there in the description. Don't miss it. All right, let's get into today's video. >> Project number one, reverse skill AI skill routing pack for security work. Reverse skill is an open-source routing pack for code AI clients like Claude code, cursor, and incline.

01:11 It guides the AI to pick the right reverse engineering, penetration testing, or CTF method before acting. Then connects local tools and MCP servers such as Judag, IDA Pro, radare2, and Frida. It runs on your own machine and logs each finished task so the skill that keeps improving. It suits security researchers doing authorized work. Add it to your AI client and route the right skill.

01:36 Project number two, chewicker terminal code review with Vim key bindings. Chewicker is a command-line code review tool that runs entirely in your terminal with Vim key bindings. It shows a continuous GitHub-style diff where you scroll through changed files and add comments at the line, range, file, or review level. It works with Git, JJ, and Mercurial.

01:57 Reviews uncommitted changes, commit ranges, or any pull request. When done, you push a real review to GitHub or GitLab, copy structured markdown for a coding agent, or pipe to state to doubt. It suits developers who prefer the keyboard. Install it and review your next change without leaving the terminal. Project number three, ESP32-Bit-Pirate, multi-protocol hardware tool on ESP32.

02:22 ESP32-Bit-Pirate is open-source firmware an ESP32-S3 board into a multi-protocol development and analysis tool inspired by the Bus Pirate. You control it through a command line interface over USB serial or a Wi-Fi web browser, and it sniffs, sends, and scripts protocols like I2C, UART, SPI, and one wire, plus radio such as Bluetooth, Wi-Fi, sub-GHz, and RFID.

02:45 You flash it from the browser, then automate work with Bus Pirate bytecode or Python. It suits hardware developers and hobbyists testing they own. Flash a supported board and start exploring protocols. Project number four, skills for real engineers, agent skills for disciplined software work. Skills for real engineers is an open-source collection of agent skills for coding tools like Claude Code and Codex.

03:10 It fixes common failure modes where an agent misunderstands the task, writes verbose code, or ships a messy design. Each skill is small, composable, and works with any model. You install them as Claude Code plugin or copy editable files into your project with skills. Shh. Then invoke them to grill your plan, build a shared project language, drive test-driven development, and review code.

03:35 It suits engineers who want structure without losing control. Add the skills to your agent and ship with more discipline. Project number five, T3 Code, control coding agents from any device. T3 Code is an open-source control surface for the coding agents already running on your machine. It works with your existing subscriptions to Claude code, Codex, Cursor, Grok Build, and Open Code, and lets you drive them from a mobile app, web app, or electron desktop app.

04:01 It runs a local back end on your computer, so the agents stay on your hardware while you steer them remotely. You start it with a single command or install the desktop app through Winget, Homebrew, or the AUR. It suits developers who want a performant, remote-ready way to manage agents. Run it and control your agents from anywhere. Project number six, Project ARI, self-hosted AI companion you fully own.

04:24 Project ARI is an open-source, self-hosted AI companion, a digital character you own and run yourself. It gives a large language model a face and a voice through Live 2D or VRM avatars, and it holds real-time voice chat with you. It connects to many model providers such as OpenAI, a llama, and vLLM, and it can chat in Discord and Telegram, or play games like Minecraft and Factorio.

04:50 The browser version uses web technologies like WebGPU and WebAssembly, while the desktop app taps native CUDA and Metal, and it runs on the web, macOS, and Windows. It suits developers and enthusiasts who want their own cyber living being. Run it and bring your companion to life. Project number seven, Skills for Design Engineers, agent skills for polished UI motion.

05:13 Skills for Design Engineers is an open-source set of agent skills that help coding tools build better interfaces and animations. It targets a common gap where agents pick the wrong easing, borders, or shadows, and it encodes design engineering judgment, so the agent chooses well. You install it with a single command through Skills. Shh. Then reach for skills that review your animations, audit a code base for motion problems, find where motion actually helps, and pick trustworthy UI libraries.

05:42 It suits designers and front-end engineers who want interfaces that feel considered. Add it to your agent and sharpen your UI motion. Project number eight, system prompts leaks, archive of AI chatbot system prompts. System prompts leaks is an open collection of extracted system prompts, the hidden instructions that AI chatbots receive before your first message.

06:01 It gathers these prompts verbatim as plain markdown files sorted into folders by company such as Anthropic, OpenAI, Google, xAI, Microsoft, Perplexity, and Mistral. Alongside the chat models, it captures the instructions behind coding agents, browser assistants, and injected mid-conversation reminders. And it also mirrors officially published prompts for comparison.

06:25 It suits developers, researchers, and prompt engineers who want to study how these systems are steered. Browse the folders and read how your favorite assistant is instructed. Project number nine, CapRover, self-hosted app deployment platform on Docker. CapRover is an open-source, self-hosted platform as a service for deploying apps, databases, and websites to your own server.

06:46 It removes the manual setup of servers and SSL, letting you ship Node.js, Python, PHP, Ruby, Go, and .NET apps without knowing Docker or Nginx directly. Under the hood, it runs Docker Swarm for containers and clustering, Nginx for load balancing, and Let's Encrypt for free HTTPS. And you can install databases like MySQL, MongoDB, and Postgres from a drop-down.

07:12 You control it through a web GUI or a CLI for automation. And it stays lock-in free, so your apps keep running if you remove it. It suits developers who want an affordable alternative to hosted services. Point it at a server and deploy your first app. Project number 10, DyanD Santa, binary authorization system for macOS. Santa is an open-source binary and file access authorization system for macOS that decides which programs are allowed to run.

07:38 It installs a system extension that watches every execution and checks it against a local database, a GUI agent that alerts you when something is blocked, and a command line tool for managing rules. In monitor mode, it logs everything while allowing most binaries, and in lockdown mode, only approved ones run. It writes rules by code signature, certificate, team ID, or file path, so you can trust a whole publisher and still block one app.

08:03 A background service syncs policies with a management server using open source options like Marauro, Rudolf, or Zentric. It suits security teams protecting fleets of Macs. Deploy it and control what runs across your machines. Project number 11, Temporal Python SDK Samples. Runnable examples for durable Python workflows. Temporal Python SDK Samples is an open source collection of runnable examples for building durable workflows with the Temporal Python SDK.

08:30 It shows how to write workflows and activities and run them on a local Temporal server, covering core patterns like signals, queries, updates, child workflows, retries, cancellation schedules, and cron. Each sample lives in its own folder, and you install dependencies with UV and run any example with a single command. It also includes integrations with tools like Open Telemetry, Prometheus, Sentry, Pydantic, LangChain, LangGraph, and Amazon Bedrock.

08:59 It suits Python developers learning how to make long-running processes reliable. Clone it and run a sample to see durable execution work. Project number 12, Codex Security. AI-powered security scanner for your code. Codex Security is an open source CLI and TypeScript SDK that finds, validates, and reviews security issues in code you own or have permission to assess.

09:22 You install it with NPM, sign in with an OpenAI account or API key, then point it at a repository, a diff, or specific paths. It runs an AI model to rank files, review them, validate findings, and trace attack paths, and it can feed your own threat models as knowledge bases. You export results as SARIF, JSON, or CSV, add a pre-commit hook, or run bulk scans across many repositories inside a hardened Docker sandbox.

09:49 It suits developers and security teams reviewing their own code bases. Install it and scan a repository you control. Project number 13, Neon, compiler-driven React alternative without virtual DOM. Octane is an open-source TypeScript-first UI framework built as a fast alternative to React. It keeps the React API you already know, so use state, use effect, memo, context, and suspense work the same, but a compiler removes the virtual DOM and much of the manual bookkeeping before your app ships.

10:21 You can omit dependency arrays and let the compiler infer them, place hooks behind conditions, and opt into a .tcrx dialect with template directives like @if and @for. It runs standard JSX out of the box, plugs into Vite, esbuild, or esbuild, and supports streaming server rendering with byte-stable hydration. It suits React developers who want the same model with less overhead.

10:44 Add it to a Vite app and build with familiar hooks. Project number 14, Playwright, cross-browser web testing and automation framework. Playwright is an open-source framework for web testing and automation that drives Chromium, Firefox, and WebKit through a single API. You use it to write end-to-end tests, run browser automation scripts, or give AI agents control of a browser.

11:05 It waits for elements automatically, uses resilient locators that mirror how users see a page, and isolates each test in its own browser context. It captures traces, screenshots, and videos to debug failures, and it runs headless or headed on Linux, macOS, and Windows. Beyond JavaScript, it works in Python, .NET, and Java, and ships a CLI, an MCP server, and a VS Code extension.

11:29 It suits developers and QA engineers testing web apps. Install it and write your first cross-browser test. Project number 15, Blaze Diff, high-performance diff toolkit for images and data. Blaze Diff is an open-source ecosystem of diff tools for JavaScript and beyond. Built for speed. At its core, it compares images pixel by pixel with a Rust and SIMD engine, plus a WebAssembly build for browsers and edge runtimes.

11:55 Around that, it adds object diffing with change tracking, perceptual quality metrics like SSIM and GMSD, a from-scratch PNG codec, React components, and a framework-agnostic renderer for diff UIs. You install packages from NPM, JSR, Cargo, or Pip. Run comparisons through a CLI or use an agent that captures screenshots and hands ambiguous diffs to your coding agent to judge.

12:20 It suits developers doing visual regression testing and data comparison. Install a package and start detecting differences. Project number 16, Webiny, self-hosted serverless CMS platform on AWS. Webiny is an open-source self-hosted content management platform that runs on AWS serverless inside your own account. It is a TypeScript framework you extend with code rather than a closed product you configure.

12:45 And it provisions its infrastructure on Lambda, DynamoDB, S3, and CloudFront through Pulumi with a single deploy command. It ships a headless CMS with a GraphQL API, a drag-and-drop website builder, a file manager, publishing workflows, and native multi-tenancy from one deployment. You add custom content models, life cycle hooks, and admin UI in the extensions folder.

13:08 And a built-in MCP server gives AI coding agents real context about the framework. It suits developers at large organizations who want data ownership and real extensibility. Create a project and deploy your own CMS to AWS. Project number 17, Nash XY, GPU-accelerated Python charts for huge data sets. XY is an open-source Python charting library for the web, notebooks, and static exports, built to handle very large data sets.

13:36 You build a chart from a container and the marks inside it using a declarative API or familiar matplotlib conventions, and customize every layer with Python, CSS, or Tailwind. For small charts, it sends every point to the browser, but for large ones, a Rust core keeps the exact values in Python and computes only what the screen needs, drawing a density surface that drills back to real rows as you pan and zoom.

14:01 You export charts as HTML, PNG, SVG, or PDF and embed them in a Reflex app through a dedicated adapter. It suits data scientists and Python developers who plot millions of points. Install it and chart a large data set interactively. Project number 18, Telivy. Self-hosted revenue-first analytics for founders. Telivy is the open-source self-hosted edition of a revenue-first analytics platform built for founders.

14:28 It combines web analytics, session replay, and revenue attribution, so you can see not just traffic, but which visits turn into paying customers. You connect customer revenue from Stripe, Lemon Squeezy, Polar, Dodo, Yolphy, or a manual payment API, and it tracks subscription life cycle, refunds, disputes, and first and last touch attribution. It runs on Node.js with a Postgres database, installs through Docker or local development, and drops a tracking snippet into your site.

14:56 It supports shared analytics, website collaborators, and import and export. It suits indie founders and small teams who want to own their data. Deploy it and connect your payments to see revenue behind your traffic. Project number 19, Interior Dev. Copy-paste React micro-interaction components. Interior Dev is an open-source collection of React micro-interaction components focused on the half second right after a click.

15:23 There is no package to install. Each component is a single file you copy into your project. Every file exports two things: a headless hook that owns all the behavior and touches no class names, and a styled component built on that hook as an example you can keep or replace. Because the logic lives in the hook, reskinning a component to your own design costs nothing but classes, and the only dependency is motion.

15:45 Each one reserves space for every state, handles keyboard input fully, and respects reduced motion settings. It suits front-end developers who want interactions that feel finished. Copy a component into your app and refine the details. Project number 20, AISD BA. Go STBA for streaming. Tool calling AI back-ends. AISD BA is Grafana's open-source Go STBA for building AI-powered back-ends and agents.

16:13 It lets a Go service generate or stream text from a language model calling Go functions as tools, and returns schema-validated structured output. It speaks the same wire protocol as the AISD BA for React, so a Go back-end can serve existing use chat, use completion, and use object front-ends without a protocol adapter. It calls Anthropic, OpenAI, Amazon Bedrock, any OpenAI compatible endpoint, or Grafana's hosted endpoint, and adds production controls like time-outs, fall-back, retries, structured logging, and

16:47 Prometheus metrics. Tools can require approval before consequential actions, and agent loops run until the task finishes. It suits Go developers building agent runtimes and AI services. Install the module and make your first model call. Thanks for watching. See you in the next update.