RxDB

We've been building ClarityBoss for a few months now, and made an early bet on a technology neither Matt Miner nor myself had used before. So far, it's been a pretty good bet! RxDB has let us build a browser-based application that has the ability to work offline, search and load data quickly since it can all happen client side, and keep changes instantly synced between multiple devices. It also sets us up nicely in the future for some privacy-focused features we are exploring. I've wrote up two semi-technical posts on some of the things we've figured out along the way- links to Part 1 (frontend concerns) and Part 2 (our backend Golang sync implementation) in the comments.

In the future, users won't browse the app stores to find the "best app" for their use case. Instead, their agent will fork the best-fitting project from GitHub and adapt it to their needs. With the latest RxDB release, we are taking a step into this direction: Users can now sync directly with their own Google Drive or Microsoft OneDrive account. No Servers. No Backend. Just user-owned data.

Just published expo-opfs. If you're building an Expo app and want to use a web library that needs the Origin Private File System (OPFS), this polyfill based on expo-filesystem makes it work out of the box. https://lnkd.in/g5Ggi-j8

The end of AI DOM scraping is here. Introducing the RxDB WebMCP Plugin! For years, automation and AI interactions in the browser have relied on a "pixels-as-APIs" approach: guessing CSS class names, parsing accessibility trees, and breaking whenever a layout changes. WebMCP (Web Model Context Protocol), currently incubated by the W3C Web Machine Learning community group, changes everything. It transforms the browser from a visual document viewer into a semantic capability surface, providing a formalized machine interface alongside the human interface. But WebMCP is uniquely powerful when paired with Local-First databases like RxDB: - Unlimited Options: Instead of hardcoding tools for every possible action, exposing a local database gives AI Agents unlimited generic query and mutation options mathematically bound only by your schema. - Zero Latency & Offline: Agents query data instantly from the local database on the user's device, even without an internet connection. - Privacy First: Sensitive user data stays on the device while still being queryable by an on-device AI model. - LLM-Friendly NoSQL: Writing JSON-based NoSQL queries is much easier for LLMs than orchestrating complex SQL JOINs. - Native JSONSchema: RxDB schemas are already written in JSONSchema, meaning zero translation overhead for WebMCP! With the new RxDB-WebMCP plugin, you can expose your RxDB collections to AI Agents with a single function call. Check out the documentation to learn how you can make your Local-First applications natively accessible to AI today: https://lnkd.in/diF9_T6t #RxDB #LocalFirst #WebMCP #AI #WebDevelopment #JavaScript #Database #LLM #OpenSource

TanStack Start is a new full-stack framework for React that's an alternative to options like Next.js and Remix. And in this tutorial, Andrew walks you through how to get started with it. You'll use it to build a CRUD app along with TanStackDB and add RxDB integration, too. https://lnkd.in/gYkbwMZG

Tenho desenvolvido um projeto offline / local-first, durante os últimos 3 meses, para o meu pai, que é fazendeiro. A necessidade era simples e real: registrar informações no curral ou no meio do pasto, onde não existe internet. Esse se tornou, facilmente, o projeto mais desafiador que já fiz. Não só pela parte técnica, mas por lidar com regras de negócio reais, vindas de alguém exigente (e que realmente usa o sistema 😅). Isso mudou totalmente minha forma de pensar como dev. Pesquisei bastante sobre arquitetura local-first, usei a IA como parceira para explorar soluções dentro do que eu já dominava, prototipei tudo no Figma e defini uma arquitetura com React/Next como PWA, banco local e sincronização com um banco global (Supabase/PostgreSQL). Comecei com SQLite como banco local, senti limitações e migrei para o RxDB, que tem funcionado muito bem no cenário offline. A sincronização do local com o global está sólida, o caminho inverso ainda tem desafios em múltiplos dispositivos, então implementei uma solução manual de reset do banco local enquanto evoluo isso. Hoje o app já conta com cadastro e gestão de animais, reprodução, vacinação, pesagem, importação de planilhas, relatórios em PDF e estou finalizando gestão de vendas, clientes e movimentações. Testei o app no curral, durante uma inseminação, totalmente offline: armazenou 66 dados reprodutivos e sincronizou tudo depois. Esse projeto reforçou o quanto construir para o mundo real acelera a evolução técnica e mental, da escolha da stack à arquitetura e à experiência de quem realmente usa o sistema. 👉 Você já construiu algo offline-first / local-first para um problema real ou já se desafiou com um projeto fora da zona de conforto? #LocalFirst #OfflineFirst #DesenvolvimentoDeSoftware #EngenhariaDeSoftware #WebDevelopment #FullStack #ReactJS #NextJS #PWA #ArquiteturaDeSoftware #BancoDeDados #UX #AprendizadoNaPrática #ProjetosReais #TechForGood #AgroTech

For the last 5 months, I’ve been keeping a bit of a secret. 🤫 Back in September, I found myself on the job market like so many others in tech recently. In between applications and interviews I needed an outlet, something to keep my engineering skills sharp and my mind focused. So I started building a solution to my own problem: a fitness app with structured lifting programs that automatically handles progressive overload. What began as a small project to stay productive quickly became an obsession. The "tinkering" turned into shipping code daily. I’m excited to finally share what I’ve been passionately working on: Pump It Better. It’s an offline-first fitness tracker for the web, focused on what matters: your progress, not just data entry. Native iOS and Android apps coming very soon! And now, I need a favor from my network. 🙏 I’m looking for a few "Technical Beta Testers" to help me kick the tires. I don’t just want users; I want engineers, designers, and product people who can give me the kind of brutal, honest feedback that helps a product grow. THE ASK: 1. Check out the PWA web app at https://lnkd.in/gf5f2snY 2. Try a workout or just poke around (it works fully offline!) 3. DM me your feedback. What breaks? What feels slow? What UX pattern drives you crazy? 🎁 THE REWARD: To say thanks, I'm granting a Free Lifetime Premium Membership ($50 value) to the first 20 people who send me detailed feedback. It’s a V1, so please be gentle! I know there are rough edges to smooth out. It’s built with an "Offline-First" architecture using RxDB and React, so I'm especially interested in feedback on the performance and user experience. Thank you for your support! #BuildingInPublic #SoftwareEngineering #FitnessTech #React #RxDB #BetaTesting #OpenToWork

Still processing that we took 2nd place at my first in-person hackathon!🥈 My team (Christopher Tela, Daimen Miller, and I) built CareLink for Western Developers Society’s Spark 2026 hackathon. It’s a PWA that helps nurses in remote Canadian communities track patient referrals when internet is unreliable. Right now, referrals get lost in paper forms and spreadsheets, which means people wait even longer for care they already wait months for. We built offline-first sync with RxDB/CouchDB (Couch running in Docker), an OCR scanner so nurses can photograph paper referrals instead of transcribing them, a kanban board for referral management, and a patient portal for appointment tracking. Biggest lesson: we tried to “one shot” the whole thing after what seemed like extensive planning. It worked, but barely. Would’ve saved ourselves a lot of scrambling with better structure from the start! Pretty surreal going from “let’s try this” to building something that could actually help underserved communities. 🫂 GitHub: https://lnkd.in/esvea8aX DevPost: https://lnkd.in/eyyRaTkX

In 2023, I asked ChatGPT to fix my docs. In 2024, I copied code back and forth to ChatGPT. In 2025, I'm directing an orchestra of LLMs to play in sync. This is 𝗩𝗶𝗯𝗲-𝗖𝗼𝗱𝗶𝗻𝗴 🎼💻✨ Vibe-Coding is not a meme anymore. It's how modern software is actually being built. So with the latest RxDB release, I did not just add features. I optimized RxDB for Vibe-Coding and LLMs, not against them. In other words: 𝗗𝗫-𝗳𝗼𝗿-𝗟𝗟𝗠𝘀 🧠 1️⃣ 𝗹𝗹𝗺𝘀.𝘁𝘅𝘁 Added /llms.txt so AI tools can read RxDB docs without drowning in irrelevant context. 2️⃣ 𝗘𝗿𝗿𝗼𝗿𝘀 𝗮𝗿𝗲 𝗻𝗼 𝗹𝗼𝗻𝗴𝗲𝗿 𝗱𝗲𝗮𝗱 𝗲𝗻𝗱𝘀 RxDB errors now include 𝗰𝗮𝘂𝘀𝗲, 𝗳𝗶𝘅, and 𝗱𝗼𝗰𝘀 fields so LLMs (and humans) are guided to the right solution. • A 𝗰𝗮𝘂𝘀𝗲 field describing which functionality or plugin triggered the problem • A 𝗳𝗶𝘅 field with a proposed solution • A 𝗱𝗼𝗰𝘀 field linking to the relevant documentation 3️⃣ 𝗥𝗲𝗱𝘂𝗰𝗲𝗱 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗡𝗼𝗶𝘀𝗲 Added .aiexclude, .claudeignore, and agent configs to reduce context noise and token usage. 4️⃣ 𝗔𝗴𝗲𝗻𝘁 𝗩𝗲𝗿𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻 Added AGENTS.md, agent skills and validation steps so AI-generated changes get checked, not blindly trusted. 🎼 The orchestra is warming up, and we’re still refining the score. 👉 How are you adapting your dev-tools for LLM-first workflows?