Lovable

Lovable shipped agent integrations on July 15 and I've barely opened my own apps since (and that's the entire point). You can add an MCP server to any publicly published Lovable app. Here's an example... Every company runs two narratives. The one on their website, and the one the models assemble from Reddit threads, G2 reviews, and comparison posts. Buyers read the second one first now. So I built NarrativeOS. Point it at a company, it reads the public record based on agents scraping thousands of public data sources, and hands you the briefing. Screenshot below is Jasper. Net sentiment -18 across 13 quotes from 23 sources from the last 7 days. Their differentiation is real, and buyers say so unprompted. Brand voice, team workflow, the stuff that matters once more than one person is writing. But the objection that decides it never gets there. It's $49 a month for a single seat. Their best asset sits behind a wall most buyers never climb. Here's the part that matters for builders. That Jasper briefing doesn't sit in a dashboard I remember to check. It lands inside Claude. Which quote moved the score. What shifted since the last run. Draft counter-positioning off finding two. Not a tool I open but a system I argue with. And Lovable makes this stupid easy now. You ask it to turn agent integrations on, it reads your app's own logic and proposes the scope of the server for you. OAuth by default, security-checked at publish, hosted and kept in sync with whatever you shipped last. I've done this by hand...this now takes a prompt. For fifteen years the whole job was getting somebody to open your app. Every onboarding email and activation dashboard existed to solve that one problem. MCP is the first thing I've seen that says you don't have to. The tab you spent a decade fighting for stops being the point. I wrote the full teardown of this launch, what it changes about distribution and pricing, and where the model breaks down. Full transparency, StackedGTM is very proudly a Lovable partner, and I want to be clear about the order that happened in. I was a super user for six months before we formalized anything. StackedGTM.AI, NarrativeOS, dozens of client work (project management, quotes, you name it...all apps, now accessible via MCP), all built in Lovable on nights and weekends. The partnership followed the usage. #lovablepartner

Claude Opus 5 is now available in Lovable. We pushed it hardest on the thing builders care about most: shipping complete, working apps. It's the top performer in its family, up 22% over Claude Opus 4.7 on our most demanding coding tasks, with the highest finished-app quality we've measured and noticeably less variance run to run.

Admin baddies x Lovable feature roundup: --> PII detection --> reusable authentication (one of our most requested!!) --> publishing controls --> workspace insights (my personal favorite - check what data users have connected to) --> automated security scanning (you can schedule these to your little hearts desire!) --> automated cleanup of abandoned apps AKA clear out zombie apps 🧟♀️ https://lnkd.in/e2fDuduE

We have swarms of AI agents playing capture the flag in our production infrastructure. We hide flags across our infra and ask our agents to find them. If they’re found, that is clear signal that there is a vulnerability. We wrote up how it works and why we do it: https://lnkd.in/gA98V74N

Every ad in SF says "agents agents agents." We skipped the billboard and shipped instead. MCP any Lovable app, instantly. To get started, ask Lovable to enable agent integrations. Learn more: https://lnkd.in/e4xchMNR

More visibility and control. Zero extra steps for your team. Here’s what we shipped recently for Enterprise teams: • Reusable authentication (one of our most requested features!): internal apps recognize your team automatically via SSO, Google, or email. • Workspace Insights: one view of your workspace, including ownership, publish status, cost, activity, security findings. Sorted by what needs attention first. • PII detection: spots if a project is storing or exposing PII data without proper safeguards. • Basic Security Scans run automatically, checking project configuration, database schema, row-level security policies, and exposed sensitive data. • Deep Security Scans can be triggered manually, for a full agentic scan, and can also be scheduled in advance. • Publishing controls: choose exactly who can see what you build: your workspace, a named list, or the public. Your source and chat history stay private either way. • Automatic cleanup of abandoned apps: prevent app sprawl and unattended data. Full breakdown on the blog: https://lnkd.in/gYd6XPha

You won't hear me say "vibe coding" much anymore. Lovable is now certified under AIUC-1, the leading third-party standard for AI security, safety, and reliability. This means we’ve been independently verified by Schellman after rigorous stress-testing of the Lovable platform against the risks that AI coding agents introduce to the world. We're way past vibes now. https://lnkd.in/dhTtMxuG

Most of what Lovable does to earn your trust, you'll never see: • Authentication and access controls with sensible defaults, out of the box • Secrets that stay encrypted and never surface in frontend code • Free security scanning that runs continuously as you build, with deployment gates before anything ships • AI-powered penetration testing, so you can see real, confirmed vulnerabilities in your own app And as of today, Lovable is certified by AIUC-1, the leading third-party standard for AI security, safety, and reliability. Earning this certification required stress-testing across risks like insecure code creation and exposed secrets, and we cleared the bar.